Isotopic Tracing and Numerical Modelling of Saline Groundwater Discharge Into Matola Wetlands, Mozambique

Isotopic Tracing and Numerical Modelling of Saline Groundwater Discharge into Matola Wetlands, Mozambique

Rezwana Binte Delwar

MSc Thesis WSE-GW.19-02 Student number 1047724 August 2019

Updated version October 2019

Isotopic Tracing and Numerical Modelling of Saline Groundwater Discharge into Matola Wetlands, Mozambique

Master of Science Thesis by Rezwana Binte Delwar

Supervisor Prof. Michael McClain, PhD, MSc (IHE Delft)

Mentors Dr. Tibor Yvan Stigter, PhD, MSc (IHE Delft) Prof. Yangxiao Zhou, PhD, MSc (IHE Delft)

Examination committee Prof. Michael McClain, PhD, MSc (IHE Delft) Prof. Yangxiao Zhou, PhD, MSc (IHE Delft) Dr. Tibor Yvan Stigter, PhD, MSc (IHE Delft) Dr. Matthijs Bonte, PhD, MSc (Shell)

This thesis is submitted in partial fulfilment of the requirements for the academic degree of Master of Science in Water Science and Engineering IHE Delft Institute for Water Education, Delft, the Netherlands Master of Science in Environmental Engineering Instituto Superior Técnico, Universidade de Lisboa, Portugal Master of Science in Hydro Science and Engineering Technische Universität Dresden, Germany

MSc research host institution IHE-Delft Institute for Water Education

Delft August 2019

Although the author and IHE Delft Institute for Water Education have made every effort to ensure that the information in this thesis was correct at press time, the author and IHE Delft do not assume and hereby disclaim any liability to any party for any loss, damage, or disruption caused by errors or omissions, whether such errors or omissions result from negligence, accident, or any other cause.

Abstract

Population pressure, repeated natural disasters and sea-level rise associated with projected climate changes increase pressure on secure freshwater resources, notably on groundwater of the coastal areas such as the Great Maputo Area in Mozambique. The Matola River, located in the west of Maputo, is a perennial river that carries mostly brackish/saltwater originating from groundwater seepage and salinity makes the river water unusable. The main origin of the saline groundwater is assumed to be fossil seawater, entrapped in the silty marl and clay dominated aquitards for thousands of years. However, detailed studies about the salinity problem of coastal Maputo city, particularly of Matola River are very limited. This research focuses on the integration of regional hydrogeochemistry, isotopic analysis and groundwater flow models to trace the source and evolution of saltwater in Matola wetlands. The Piper plot, Stiff map and bivariate diagrams of major ions reveal salinization through mixing with seawater followed by ion-exchange as the prominent hydrochemical processes of the study area west of the Matola River. On the other hand, east of the Matola River the dominant process seems to be freshening, and this distribution is related to land cover connected recharge rates of the area and the hydraulic properties of the aquitards. Water stable isotopes (δ2H, δ18O) and 18O/Cl further confirm the mixing process between fresh and connate saline groundwater in conjunction with evaporation of shallow groundwater and surface water being the major sources of salinity. The advective transport model was developed in PMWIN based on the existing steady-state model to examine the residence time and flow paths of the groundwater, particularly for the area of Matola River and wetlands. The existing steady-state model shows the highest sensitivity to effective porosity, followed by hydraulic conductivities and model geometry. The range of travel times obtained through particle tracking was frequently large and highly dependent on the number of particles. Using the result of 13C/14C (DIC) isotopes, the evaluation of the existing groundwater flow model was undertaken. Despite several uncertainties in the tracer age calculations as well as model-simulated residence time, both studies agree the residence time of groundwater to be approximately 6000 years at maximum, with an average range of 3000- 4000 years. Development of two salt transport models (SEAWAT) further aided to assess the mentioned hypotheses. The “paleo” SEAWAT model partially validates the result of 14C tracer and PMPATH simulated residence time. On the other hand, the “modern” SEAWAT model predicts fresher aquifer system within next 1000-1500 years approximately. Based on the overall findings in combination with the recent transgressional history of the area, this research infers that the Holocene connate seawater cannot be the only source of salinity in Matola wetlands; other sources such as evaporite dissolution and slow diffusion of formation water (older than Holocene period) might also stimulate the intensification of the salinity problem of Matola wetlands. The results obtained in this research can contribute to target adaptation and mitigation measures for high groundwater salinity and assist the policymakers to take feasible solutions to sustainable groundwater management for Maputo region.

Acknowledgements

Foremost, I would like to thank my father for being my supporter, being the only person to believe in me and allowing me to pursue my dream. Thanks to my other family members and friends from Bangladesh for tolerating my bad mood and giving continuous inspiration. Thank you Dr. Tibor Yvan Stigter, for your curiosity, valuable advice, constructive comments and encouragement throughout this journey. Thank you Prof. Yangxiao Zhou, for your guidance, instructions and patience. I am in great indebtedness to these two of my guardian angels. I would also like to express my sincere gratitude to all the teachers and colleagues with whom my path have crossed and who helped me to grow my knowledge directly or indirectly. Thanks to Eng. Fátima Mussa, people from ARA-Sul and University of Eduardo Mondlane who provided their support during the field work in Maputo. I also want to thank IHE-Delft Laboratory staffs for their help. Special thanks to Alberto Casillas and Ahmed Ameen for allocating time to answer my doubts within their busy schedules. No words can fully express my emotions towards the GroundwatCH family for this amazing journey of past two years. All the small gestures, care, laughter, sleepless nights and craziness hold a special place to my heart. You are my second family and I wish the very best to all of you. Special gratitude to my Delft colleagues for taking care of me when I was sick during thesis phase. Heartiest love to Hafsa for all the long conversations, listening to my naggings and sticking with me from the very beginning. It is also worth mentioning her name because she allowed me to come to Delft to do this research (!!!) Each of the three cities has different stories to tell. They have seen me struggling, feeling frustrated and crying. Again they have witnessed me going beyond my comfort zones, finding confidence and laughing with joy. These three cities have contributed a lot in shaping my mind and maturity, whatever I have today. So, I am beholden to Lisbon, Dresden and Delft too. Thanks to Sandipa didi for showing sisterly affection towards me. Without you, my first few days in Delft would have been worse. And at last but not the least I would thank the person who was always with me, invisibly. Sometimes few is more.

নয়ন তেোমোরে পোয় নো তেখিরে, েরয়ছ নয়রন নয়রন।

Rezwana Delft 26th August, 2019

iii

Table of Contents

Abstract ...... i

Acknowledgements ...... ii

Table of Contents ...... iv

List of Figures ...... vi

List of Tables ...... viii

Abbreviations ...... ix

Introduction ...... 1

1.1 Background ...... 1 1.2 Problem Statement ...... 2 1.3 Hypothesis and research objectives ...... 3 1.4 Research Questions ...... 3 State of Art ...... 4

2.1 Sources and hydrogeochemistry of saline groundwater ...... 4 2.2 Contribution of environmental isotopes in studying groundwater salinity and residence time 6 2.2.1 Environmental stable isotopes ...... 6 2.2.2 Radioactive isotopes ...... 6 2.3 Groundwater flow models ...... 8 Study Area Description ...... 10

3.1 General overview ...... 10 3.2 Climate ...... 11 3.3 Hydrology ...... 11 3.4 Regional geology ...... 11 3.5 Hydrogeology ...... 12 Research Methodologies ...... 16

4.1 Data collection and sampling ...... 17 4.1.1 Preliminary data collection ...... 17 4.1.2 Well inventory ...... 17

4.1.3 In situ measurements ...... 18 4.1.4 Laboratory analysis ...... 19 4.2 Hydrochemical analysis ...... 19 4.3 Radioactive Isotope analysis ...... 20 4.3.1 Hydrochemical Model ...... 20 4.3.2 Graphical method ...... 21 4.4 Groundwater Flow Model Development ...... 22 4.4.1 Evaluation of existing steady-state flow model ...... 23 4.4.2 Particle tracking flow path analysis ...... 24 4.4.3 Salinisation Evolution Model (SEAWAT) ...... 25 Results ...... 28

5.1 Hydrochemistry ...... 28 5.1.1 Water types ...... 28 5.1.2 Major ions trend ...... 30 5.1.3 Saturation Indices ...... 34 5.1.4 Water stable isotopic analysis ...... 35 5.2 Groundwater flow model ...... 37 5.2.1 Steady-state Model Re-adjustment ...... 37 5.2.2 Steady-state Model Sensitivity Analysis ...... 39 5.2.3 SEAWAT Model Simulations ...... 42 5.3 Radiocarbon isotopic analysis (14C age estimation) ...... 49 Discussion ...... 53

6.1 Conceptualisation and steady-state model of salinisation ...... 53 6.2 Residence time of groundwater ...... 57 6.2.1 Model simulated residence time ...... 57 6.2.2 14C calculated apparent age ...... 58 6.2.3 Comparison of model-simulated and 14C calculated transit time ...... 60 6.3 Salinisation evolution ...... 63 6.4 Salinisation Sources ...... 63 Conclusions and Recommendations ...... 68

References ...... 71

Annexes ...... 79

List of Figures

Figure 1.1-1 General overview of threats to a coastal aquifer: changes (e.g. urbanisation, sea- level rise) and consequences (e.g. subsidence, salinisation); after (Delsman et al., 2014). Source: (Hung Van Pham et al., 2019) ...... 2 Figure 2.1-1 Lateral lithostratigraphic changes during marine transgression and regression phase, Source: (Brooks Cole Thomson Learning | Open Library ) ...... 5 Figure 3.1-1 Geographical location and elevation of the study area ...... 10 Figure 3.4-1 shows the surface geological groups, whereas the dotted boundary indicates an approximated study area of this research...... 12 Figure 3.4-2 Regional geology of the Greater Maputo, Source: (Nogueira, 2017) ...... 12 Figure 3.5-1 West - East Cross-section on the southern part of Maputo City covering Matola River, Source: (WE Consult, 2013) ...... 13 Figure 3.5-2 Groundwater head difference map between the phreatic and semi-confined aquifers. Areas in blue or positive values represent zones where groundwater flow downwards (phreactic level is higher than piezometric level); areas in red or negative values represent upward flow (phreactic level is lower than piezometric level). Source : (Nogueira, 2017). ... 14 Figure 3.5-3 EC distribution of both aquifers. Blue colour = fresh groundwater; red colour = brackish. Source: (Nogueira, 2017) ...... 15 Figure 3.5-4 Flowchart of the research methodologies ...... 16 Figure 4.1-1 Distribution of collected water samples during the field visit ...... 17 Figure 4.1-2 Examples of ARA-Sul monitoring well with two piezometers (left); and private/domestic well (right) ...... 18 Figure 4.4-1 The overall steps followed to determine the residence time of groundwater ...... 22 Figure 4.4-2 The overall steps followed to determine the origin and evolution of saltwater .. 22 Figure 4.4-3 The distribution of flow paths and age of groundwater seepage into a stream, adopted from Modica et al. (1998) ...... 23 Figure 4.4-4 PMPATH settings; a) 6 faces of a cell; b) 48 particles in a cell of phreatic aquifer; c) 96 particles in a cell of semi-confined aquifer; d) colours provided in different layers, where blue represents phreatic aquifer, orange represents aquitard, cyan represents semi-confined aquifer and yellow represents bottom aquitard...... 25 Figure 5.1-1 Stiff diagrams of the samples where red polygons are piezometer wells, granular polygons are surface water, pale green polygon is auger-hole water and striped polygons are domestic (private) wells...... 29 Figure 5.1-2 Piper plot of the collected samples from Matola wetlands, the red line indicates saltwater intrusion, the blue line indicates freshening, black line indicates mixing (after Han et al., 2011)...... 30 Figure 5.1-3 Scatter plot of major ions a) Na-Cl, b) Ca-HCO3, c) Ca-Cl, d) Ca-SO4, e) HCO3- Cl, f) SO4-Cl; g) SO4-Fe and h) pCO2-HCO3 ...... 32 Figure 5.1-4 Scatter of major ion ratios a) Na/Cl-Cl, b) SO4/Cl-Cl, and c) Ca/HCO3-Cl...... 33 Figure 5.1-5 Saturation indices of major minerals ...... 34 Figure 5.1-6 δ 18O vs δ 2H plot ...... 35 Figure 5.1-7 δ18O vs Cl plot ...... 36 Figure 5.2-1 Comparison of the hydraulic head after calibrating the model with the changed drain elevation for both phreatic and semi-confined aquifers ...... 37

Figure 5.2-2 a) Particle paths of points at initial condition; b) Particles paths after re-adjustment; where number increases from upstream to downstream; blue colour represents phreatic aquifer, orange colour represents aquitard, cyan colour represents semi-confined aquifer, and yellow colour represents bottom aquitard...... 38 Figure 5.2-3 Residence times at different distances from upstream to downstream point of the Matola River ...... 39 Figure 5.2-4 Upper and lower boundaries of the modelled residence times from the manual sensitivity analysis regarding effective porosity and hydraulic conductivities ...... 40 Figure 5.2-5 Upper and lower boundaries of the modelled residence times from the manual sensitivity analysis regarding particle numbers ...... 41 Figure 5.2-6 Concentration-time curve of different observation points over 5000 years ...... 43 Figure 5.2-7 Paleo SEAWAT model simulation results of the phreatic aquifer with different effective porosity (ne) in layer 2 and layer 4 ...... 44 Figure 5.2-8 Paleo SEAWAT model simulation results of the semi-confined aquifer with different effective porosity (ne) in layer 2 and layer ...... 45 Figure 5.2-9 Concentration-time curve of different observation points over 5000 years ...... 47 Figure 5.2-10 Present SEAWAT model results of the phreatic aquifer (top row) and the semi- confined aquifer (bottom row) with different effective porosity (ne) of layer 2 and layer 4 ... 48 Figure 5.3-1 Average initial activity (as pMC) by each model for water samples ...... 50 Figure 5.3-2. a) Graphical presentation of data from Matola catchment. Sample numbers are according to Table 5.3-1; b) Han and Plummer plot from Han and Plummer, (2016)...... 51 Figure 6.1-1 Conceptual model developed by (Nogueira, 2017)and improved by (Trasviña, 2018) ...... 53 Figure 6.1-2 (a) Recharge potential rates as input in the numerical model, (b) Land cover studied by (Andreetta, 2018) ...... 56 Figure 6.1-3 Horizontal hydraulic conductivities as input in the numerical model ...... 56 Figure 6.1-4 Layer thickness in the numerical model with the location of 4 newly collected samples (column view) ...... 57 Figure 6.2-1 Flownet of Matola 1 in the model with local and regional flow paths ...... 58 Figure 6.2-2 Simplified plot of the changes in 14C age, DIC concentration and δ13C (DIC) values in groundwater as a result of fossil organic matter oxidation and the successive fossil carbonate dissolution. Source: Geyh (2000) ...... 60 Figure 6.2-3 Correlation between the hydrochemical models calculated and steady-state model- simulated residence times; (a) Pearson model vs simulated mean RT, (b) Pearson model vs simulated median RT, (c) Han and Plummer model vs simulated mean RT, (d) Han and Plummer model vs simulated median RT and the black dotted line in 1:1 line...... 61 Figure 6.2-4 Comparison of residence times after model simulation and hydrochemical model calculation ...... 62 Figure 6.4-1 Holocene sea-level fluctuations in southern Africa based on beach rock dating, Source: (Ramsay, 1995) ...... 64 Figure 6.4-2 (a) Marine transgression of Holocene period (up to 4 m), (b) Relative elevation of sea-level needed to inundate the sampling points (60 m) ...... 65 Figure 6.4-3 Inferred positions of paleogeographic coastlines in southern Mozambique, red box indicates the present location of Matola River; Source: (Förster, 1975) ...... 67

vii

List of Tables

Table 2.2-1 Summary of geochemical reactions included in selected traditional Ao adjustment models ...... 7 Table 4.3-1 Presumed parameter values used in the hydrochemical models ...... 20 Table 4.4-1 Simulation of different hydraulic parameter values of the aquitards in different sectors of the aquifer for manual sensitivity analysis ...... 24 Table 4.4-2 Initial dispersion values ...... 26 Table 4.4-3 Effective porosity values of different layers ...... 26 Table 4.4-4 Sensitivity analysis of SEAWAT model concerning effective porosity ...... 27 Table 5.2-1 Water budget for steady-state models ...... 38 Table 5.2-2 Comparison of residence times after re-adjustment (along distance from upstream) ...... 39 Table 5.2-3 Summary of the manual sensitivity analysis of the steady-state model...... 40 Table 5.2-4 Summary of descriptive statistics of the simulated mean residence times (in years) at different observation points ...... 41 Table 5.2-5 Qualitative summary of the Paleo SEAWAT model simulation ...... 42 Table 5.3-1 Isotopic data of the sampling sites; ...... 49 Table 5.3-2 Calculated initial activity (as pMC) of water samples, using six hydrochemical models...... 49 Table 5.3-3 Ages of water samples (in years BP) calculated using six hydrochemical models ...... 52 Table 6.2-1 Groundwater residence time determined by PMPATH and apparent age determined by 14C tracer; where “C” = contaminated or modern sample having negative values in the calculation ...... 62

viii

Abbreviations

ARA-Sul Administração Regional de Aguas do Sul PZ Piezometer M Private/domestic wells (boreholes) F Phreatic aquifer SC Semi-confined aquifer BEX Base Exchange Index ICP-MS Inductively coupled plasma mass spectrometry IC Ion Chromatography VPDB Vienna Peedee Belemnite IRMS Isotope-ratio Mass Spectrometry AMS Accelerator Mass Spectrometry pMC Percent Modern Carbon TTD Transit Time Distribution RT Residence Time

Introduction

This chapter proposes the motivation of this research. It presents the background of groundwater salinisation problem in the study area, the hypotheses based research objectives and the questions that this study will discuss.

1.1 Background

Groundwater, which is found under the earth’s surface, holds 96% of the earth’s liquid freshwater. It is recognised as a critical freshwater resource for human survival and as well as socio-economic development (Shiklomanov and Rodda, 2003). Above 2 billion people across world are dependent on groundwater on daily basis (Kemper, 2004). Groundwater is often chosen in developing countries because it has better quality and thus does not require much treatment (Appelo, and Postma, 2005). Presently 150 km of the coasts are inhabited by about 44% of the world’s population (UN Atlas of the Oceans, 2019). Increasing demand along the coasts due to population explosion along with climate change (e.g. more extended drought period) has, and will put pressure on water resources in future years (Werner and Simmons, 2009; Treidel et al., 2011). However, coastal groundwater is considerably sensitive to many factors such as climatic variations (e.g. (Stigter et al., 2014, 2017); seawater intrusion due to intense aquifer exploitation (Boukhari, et al., 2015; Han, et al.,, 2014; Kim et al., 2003; Werner & Simmons, 2009); upconing of deep saline paleowaters and its evolution in the coastal wetlands through mixing with fresh recharge water; water-rock interaction and evaporation (Han, et al., 2011; Vengosh et al., 2005; Sola, et al., 2014); dissolution of evaporites (Mongelli et al., 2013; Cendón et al., 2019); pollution by untreated wastewater (Ayadi et al., 2018), and agricultural return flow (Stigter et al., 1998; 2006; 2008) among others. Since these processes are not necessarily unique and often act together, thus defining the origin of salinisation might become highly complicated. As a consequence, diversified threats on ecology, socio-economy, and stakeholders can be triggered. Deterioration of groundwater quality (Han et al., 2014; Kim et al., 2003), land subsidence (Xu et al., 2008), degradation of wetland ecosystems and groundwater dependent biodiversity loss are some of the few reasons why coastal groundwater has earned much consideration in the 21st century. Figure 1.1-1 explains the changes and consequences of coastal aquifers.

Figure 1.1-1 General overview of threats to a coastal aquifer: changes (e.g. urbanisation, sea-level rise) and consequences (e.g. subsidence, salinisation); after (Delsman et al., 2014). Source: (Hung Van Pham et al., 2019) Matola River and its wetlands in southern Mozambique are the focused areas of this research study. Matola is the largest suburb of the capital, Maputo. Groundwater is the primary source of drinking water for this region (Cendón et al., 2019). Since the mid-1980’s vast numbers of bores were dug. As a result, independent groundwater small scale providers became the reliable sources of supply water in many parts of the metropolitan (Cendón et al., 2019). Currently, groundwater is an inseparable part of the lives of this Indian Ocean port city inhabitants. Secure water resources, especially groundwater, is under great threat due to population pressure, urbanisation, recent droughts and sea-level rise induced by climate change (NCEA, 2015). The local water agency, Southern Regional Water Authority (ARA-Sul) manages groundwater resources by monitoring programs on a network of bores. Within seven years (2008-2014), approximately 1200 boreholes were increased to supply water domestically in Maputo (Nogueira, 2017). But the local aquifer is contaminated by the brackish to saline groundwater which inhibit the potentiality of groundwater use. However, insufficient quality information is at hand to understand the origin, saline-freshwater interactions, hydrochemical or isotopic studies to characterise groundwater residence times and vulnerabilities-adaptation measures of groundwater resources of Maputo (Nogueira, 2017; Cendón et al., 2019). Recently, the DUPC2 SALINPROVE project, set up by IHE Delft in collaboration with the University of Eduardo Mondlane in Maputo and ARA-Sul is assessing the occurrence of groundwater salinities, possible impacts and proposing adaptation and mitigation solutions for the Great Maputo region. The present research is the continuation of aiming for the goals of the project.

1.2 Problem Statement

The brackish/salty Matola River water greatly hampers the domestic water supply and crop productivity sectors of the adjacent areas. To correctly target sustainable mitigation or adaptation measures for groundwater salinisation, it is crucial to distinguish and understand the

prominent salinisation mechanisms that occur and their link to the conceptual models of groundwater flow. The sources of salinity observed in Matola River have been studied through hydrochemical analysis (Nogueira, 2017), numerical simulations (Trasviña, 2018), and environmental isotopic techniques (Cendón et al., 2019). However, there is no published work linking modelled residence time with the apparent age of groundwater and integrating the regional hydrogeochemistry, particularly for Matola River. This combined study will give a clear and detailed vision of the hydrogeological and hydrochemical evolution of the groundwater of Matola wetlands.

1.3 Hypothesis and research objectives

Two leading hypotheses have initiated the primary objective of this study: 1) Matola River is gaining water from saline groundwater seepage flowing upwards from the lower semi-confined tertiary aquifer through the silt-marl dominated aquitard. 2) The saline groundwater has a residence time of thousands of years and originates from mixing with seawater trapped in the aquitard during higher sea levels in the past.

This research aims to test the hypothesis and thereby incorporate the numerical models, environmental isotopes and hydrochemistry as tools to trace the origin and flow paths of Matola wetlands. For achieving this final objective, this study conforms several specific objectives: 1) Evaluate seawater entrapment, transgression periods, groundwater flow dynamics and saltwater/freshwater interactions of the aquifer using hydrochemistry and environmental isotopes (18O, 2H, 14C, and 13C). 2) Perform a sensitivity analysis of previously developed steady-state model to simulate the groundwater residence time and to couple with apparent isotopic age. 3) Simulate saltwater transport model to evaluate paleo and to predict future conditions of the aquifer regarding salinity. These objectives consequently will help to perceive the geological and hydrogeological setting of the greater Maputo aquifer.

1.4 Research Questions

To understand the salinisation mechanism of Matola River and its adjacent wetland, this research will look for answers of the following questions: 1. What is(are) the uncertainty(ies) in the conceptual and numerical models regarding residence time, flow paths and origin of saline groundwater in the Matola aquifers? 2. To what extent does the use of environmental isotopes, in particular, 14C, contribute to improving the conceptual and numerical models?

State of Art

This chapter provides a literature overview of different possible salinisation mechanisms and hydrochemical tools to assess them. It consists of three parts. The first part concentrates on sources and hydrogeochemistry of groundwater salinisation. The second section summarises the attributes of environmental isotopes regarding groundwater residence time and third part comprises the use of groundwater flow models assessing the groundwater salinity.

2.1 Sources and hydrogeochemistry of saline groundwater

There can be both natural and manmade sources for the presence of saltwater in the aquifer. Seawater intrusion, temperature and precipitation inconsistency due to climate change and geological aspect such as- subsidence and uplift, drive the advance and retreat of the freshwater/ saltwater interface (Carol, Kruse and Mas-Pla, 2009). Residual or fossil saline water is the natural saltwater entrapped on the low-lying sedimentary layers of coastal aquifers during paleo sea-level high stand. (Santucci et al., 2016). Hence, several marine transgression and regression phases, taking place from late Pleistocene (12 ka) and onwards, are controlling mechanisms for the development of Quaternary coastal aquifers around the world (Petalas and Diamantis, 1999; Delsman et al., 2014; Hung Van Pham et al., 2019). Extensive saline groundwater has been observed even 100 km inland Quaternary aquifers (Larsen et al., 2017). During the phases of transgressions, fine-grained marine sediments, rich in organic matter, clays, silts and fine sands, are deposited. While in regressions phases, coarse-grained alluvial-fluvial deposits dominate the deposition. These geologic changes lead to the development of multi-aquifer systems. In those systems, high-permeable alluvial and fluvial deposits form aquifers, and low-permeable marine deposits form interlude aquitards with residual saltwater (Larsen et al., 2017) as illustrated in Figure 2.1-1.

Transgression Regression Figure 2.1-1 Lateral lithostratigraphic changes during marine transgression and regression phase, Source: (Brooks Cole Thomson Learning | Open Library ) Hydrogeochemical studies are of great assist to verify the presence and origin of fossil saline water in the aquifer. Interestingly, the groundwater chemistry changes not only due to geology but also to recharge conditions and land use, even in one aquifer. The distribution of groundwater major ions differs spatially also, from recharge to discharge areas, considering various mechanisms. Cation exchange, water-rock interactions, mixing, and evaporation can be mentioned as dominant ones (Freeze and Cherry, 1979; Appelo, and Postma, 2005; Treidel et al., 2011). These mechanisms result in different physicochemical characteristics of groundwater, and consequently, various water types. For example, recharge water quality evolves from acidic to basic, unstable to stable water quality and oxic to anoxic conditions when it reaches discharge areas (Stuyfzand, 1989). Physiochemical parameters of water and major ion ratios will aid in understanding these processes (Appelo and Postma, 2005). More often, the graphical representation of ion concentration data, piper plot and stiff diagrams are commonly used to determine the main characteristics of waters (Stigter et al., 1998; Carol et al., 2009; Han et al., 2011; Karroum et al., 2017). These diagrams have been used successfully for the evolution of chemical processes throughout the world and Bagheri et al., 2015; Boukhari et al., 2015 are examples of some of them. For determining the presence of saltwater in the aquifer, electrical conductivity (EC) or total dissolved solids (TDS) are the fundamental indicators. If EC is less than approximately 1500 µS/cm, the groundwater is considered as fresh (Ouhamdouch, Bahir and Carreira, 2017). Ion ratios are also used in many coastline aquifers of the world to determine the origin of salinity including Mozambique (Graas and Savenije, 2010), China (Han, et al, 2015), Morocco (Bahir, et al, 2018), Middle East (Bagheri, et al, 2017), Italy (Mongelli et al., 2013) and southern France (Le Gal La Salle et al., 2013). When chloride concentrations are plotted versus other major ions, it reveals mixing processes of fresh and saline water (Appelo and Postma, 2005). In the context of Mozambique, Cendón, et al. (2019) concerning Freitas (1962), arranges summarised lithological and hydrological data of the Great Maputo. Freitas (1962) found high total dissolved solids (TDS) in the water samples of inland Quaternary sand-dunes of Maputo.

In the mid-1980s several investigations were done by De Sonneville, 1984 and Bouman, 1985, and they summarised the overall hydrochemistry of Maputo. They also analysed semi-confined groundwater samples and reported in Piper diagrams (Cendón et al., 2018). Nogueira, et al., (2019) also examined the hydrogeochemistry and major water types of the greater Maputo area. In his study, he concluded that there are six major water types with respect to Stuyfzand classes and Matola River has higher salinity (14700-17680 µS/cm) than Incomati River (500-800 µS/cm).

2.2 Contribution of environmental isotopes in studying groundwater salinity and residence time

The movement and transformation of water within an aquifer system can be estimated directly by using isotopes. From the fluctuation in the concentration of environmental isotopes in natural groundwater, recharge settings, aquifer complex and/or dominant processes under which the groundwater has progressed can be determined (Payne, 1983). Among the different approaches, isotope techniques are notably effective for identifying the source of salinity and renewability of groundwater all over the world (Bouchaou et al., 2018; Carol et al., 2009). 2.2.1 Environmental stable isotopes The chemical composition of stable isotopes (18O, 2H) does not change within the aquifer, instead they get fractionated (Appelo and Postma, 2005). Thus stable isotopes can be an excellent indicator of the origin, recharge settings-processes and the climatic conditions during recharge (Han et al., 2011; Kim et al., 2003; Travi et al., 2008). During the evaporation of water, the stable isotopes become fractionated and as a result heavier isotopes get enriched in the source water (Appelo and Postma, 2005). Therefore, the stable isotopes of groundwater lying below the local meteoritic water line (LMWL) indicates meteoric rain water is the source of groundwater recharge and the water has infiltrated into the soil after some extent of evaporation. Additionally, mixing between different types of water leads to isotopic ratio differences rather than fractionation, which can also be distinguishable signal of source of groundwater. The behaviour of 18O vs Cl concentration is a conventional separator for diverse processes of groundwater salinisation, such as- evaporative strata development, salt dissolution, and fossil seawater entrapment. For instance, 18O is affected by evaporation more significantly than Cl, whereas dissolution of salt demonstrates the escalation in Cl only (Boukhari et al., 2015).

2.2.2 Radioactive isotopes Radioactive environmental isotopes, particularly 14C, have verified as invaluable tools for defining groundwater age or travel times. 14C is generated in the atmosphere through the 14 interaction of N2 with cosmic rays. The half-life of C is 5730 years and its applicability is up to 30 ka to trace groundwater residence times. Muennich (1957) first discussed the use of 14C in dating groundwater and subsequently 14C has been extensively applied due to the general presence of dissolved inorganic carbon (DIC) in groundwater (Atkinson et al., 2014). The calculation of 14C residence times may be complex if groundwater DIC is resultant of mixed sources. Yechieli et al., (2009) and Zhao et al., (2018) discriminated between modern seawater

intrusion and paleo seawater based on 14C measurements and found that the fossil saline groundwater age get increased with increasing depth, which is associated to Holocene sea level rises. Groundwater recharged after 1950 may have atypically high 14C values due to the atmospheric nuclear tests. Apart from this uncertainty, several models have been suggested based on both major ions and stable carbon isotope geochemistry to correct apparent 14C ages or residence times (Han & Plummer, 2013). If the 14C decay rate and the initial 14C content is known, the age of the dissolved inorganic carbon in groundwater can be calculated from the measured amount of 14C,

휏 14퐶 푡 = − 푙푛 Equ. 1 푙푛2 14퐶표

14 14 14 14 where C and C0 are the measured and initial C, τ is the half-life of C, which is 5730 ± 30 years (Godwin, 1962) and t is the groundwater age. Radiocarbon age calculations for groundwater are not as direct as organic elements (wood, peat, bone, etc.), where it is assumed that the initial activity of 14C within the organic material was 100% of modern CO2 activity (100 pMC) (Fontes and Garnier, 1979; Gallagher et al., 2000). 14 14 There are many uncertainties in resolving the initial C content of DIC ( C0) which ultimately limit the calculation of dissolved inorganic radiocarbon age within groundwater. Moreover, in addition to radioactive decay, several simultaneously enduring chemical and physical procedures change the 14C content along groundwater flow paths. The transformation of composition and concentration of carbon isotopes (DIC) in groundwater usually comprises three basic steps: (1) soil-gaseous CO2 dissolution in water; (2) isotope fractionation of water– soil gas under open-system conditions; and (3) isotope fractionation of water–soil gas under closed-system conditions. In addition to these three steps, the composition and concentration of carbon isotopes (DIC) can be altered by several other geochemical progressions (Han et al., 2012; Han and Plummer, 2013, 2016). For instance, the change in 14C age due to the oxidation of fossil organic matter was discussed by Geyh (2000).

14 14 The initial activity (Ao) is also defined as the C content ( C0) (Wigley, 1975). To estimate Ao, few of the numerous existing single sample based models (traditional adjustment models) are- 1) Vogel’s (1967) model, 2) Tamers' (1975) model, 3) Mook’s (1972) model, 4) Pearson's (1967) model, 5) Fontes and Garnier's (1979) model, 6) Eichinger's (1983) model, 7) Han and Plummer's (2013) model, 8) IAEA model. A brief overview of the models can be seen in Table 2.2-1 (adapted from IAEA, 2013):

Table 2.2-1 Summary of geochemical reactions included in selected traditional Ao adjustment models

Processes Conventional Vogel Tamers Mass Pearson IAEA Mook Evans Eichinger Fontes and 14C age balance Garnier Carbonate X X X X X X X X dissolution Soil gas CO2 X X X X X X X X dissolution CO2 gas - aqueous X X X exchange - Calcite - HCO3 X X X exchange Gypsum dissolution X Ca/Na cation X exchange

As shown in Table 2.2-1, each model presumes certain chemical reactions and isotopic fractionations that may occur all along the compositional evolution of the dissolved inorganic 14 carbon isotopes and alter the C0 activity. Successful application of these models depends on 14 13 the carbon isotopic analysis ( CDIC and δ C) and background chemical data to conceptualise 14 the probable processes which influence the CDIC content in the aquifers (Han and Plummer, 2013). Widespread utilisation of the hydrochemical models can introduce significant uncertainties in groundwater age estimation (Han and Plummer, 2016). The graphical analysis method developed by Han et al. (2012) can help to determine the most fitting approach for a groundwater system with prevailing hydrochemical conditions, which was also attempted in this study. Ultimately, coupled with proper models, the graphical method can enhance the radiocarbon age assessment. (Meredith et al., 2018). Aside from single tracer analysis, the multi tracer study (e.g. 3H, 14C, 36Cl) is also a widely used tool to inquiry mixing in aquifers (Szabo et al., 1996; Atkinson et al., 2014; Müller et al., 2016). The changes in isotope content along flowpaths demonstrate the residence time and evolution of the water, notably the mixing process, salinisation and discharge processes (Terway, 2012); (Kendall, 1998). Cendón et al., (2019) implemented the first isotopic analysis to determine groundwater travel times in the Great Maputo area with the aid of major ions, selected trace elements, water stable isotopes (18O, 2H), carbon stable isotopes (δ13CDIC, δ13CDOC), radiocarbon (14CDIC), strontium isotopes (87Sr/86Sr), and tritium (3H), though the main focus of the research was freshwater. Nogueira et al. (2019) also studied the coastal aquifer system of the same region by water stable isotopes and multivariate statistical tools for better understanding of the freshwater occurrence under prevailing factors and the sources of high salinity. From both of the studies, it can be concluded that the main culprit for the salinisation of the freshwater is not the present- day seawater intrusion, rather evaporation. Intense rainfall events control the recharge rate significantly also. However, strong evaporation and interception also play a vital aspect in low recharge areas. Contrarily, deeper aquifer salinisation is more related to mixing with entrapped (within aquitard layers) fossil seawater.

2.3 Groundwater flow models

For the conceptual understanding of the groundwater flow system and for simulating the flow numerically groundwater models (GWM) have been used as assessment tools. GWM can also be used for replicating saltwater intrusion or occurrence in the costal aquifers. In addition, natural and human induced processes can be simulated within the models to assess the present or possible future stresses on the aquifer system. GWM can further describe groundwater transit time and flow pathways using the advective transport model, PMPATH/MODPATH (Zhou and Li, 2011). Transit times help to define the water budget of a catchment as well as geochemical evolution or perseverance of solute (contaminants). When the water stays in contact with aquifer matrix for longer time, the travel times of groundwater also get higher. Accordingly, the hydrogeochemical evolution of the aquifer system can be determined through the mean transit time, or transit time distribution (McGuire and McDonnell, 2006). PMPATH is a semi-analytical particle tracking system that calculates the groundwater flowpaths and travel times (Pollock, 1989). Transit time generated by PMPATH shows the time

occupied by a groundwater parcel to travel in a volume from the inlet cell (groundwater recharge zone) to an outlet cell (groundwater discharge zone, e.g. a pumping well or a river) (Pollock, 1989; Maloszewski and Zuber, 1996). In the PMPATH, particle pathlines and related transit times are generated by velocities attained from groundwater flow models. These pathlines and transit times considers only advective transport and neglects other transport processes and parameters like dispersion, diffusion, sorption, decay, and dual porosity (IAEA, 1993). Nonetheless, the importance of matrix porosity cannot be ignored in groundwater transit time, since it is related to Darcy velocity or hydraulic conductivity (IAEA, 1993). Since both backward and forward particle tracking is possible for the steady-state and transient model, the reconstruction of marine transgression- regression phases can be simulated emphasising either on a particular area or on the whole aquifer system as well as predicting future saltwater intrusion impacts (Delsman et al., 2014). Thus, integrated studies of regional hydrogeochemistry, isotopic analysis, and groundwater flow models have been used as a useful tool to simulate isotopic tracer concentrations and tracer ages of groundwater at the discharge points and to improve the management of groundwater resources in many areas of the world. For instance, Smidt et al. (1990) used hydrochemistry and isotopic analysis and discussed some groundwater management systems of the Maputo region. Similarly, Kalin and Long (1993) studied the radiocarbon age of the Tucson basin groundwater and compared with the temporal water flow as modelled with MODFLOW. Gusyev et al., (2014) simulated tritium concentrations and groundwater transit times of the Western Lake Taupo catchment, New Zealand with particle-tracking (MODPATH) and compared those to solute transport (MT3DMS) models. Later, they compared the resulting MODPATH tritium concentrations to measured tritium concentrations and the MT3DMS- simulated tritium concentrations. Hung Van Pham et al., (2019) simulated variable-density groundwater flow and salt transport model (SEAWAT) for considering the marine transgression-regression events over the paleo 60 ka and to portray the present-day fresh-saline groundwater distribution and groundwater ages of Vietnamese Mekong Delta. To assess the potential for decreased water levels and discharge to streams of the Maputo basin, both conceptual and numerical models of groundwater flow in the coastal aquifers were developed. Trasviña (2018) developed a steady-state model based on the conceptual model and data analysed by (Nogueira, 2017). In the same research, he presented a regional groundwater flow, MODFLOW and transport models, SEAWAT (MT3DMS), followed by multiple simulated scenarios to depict hydrological conditions and impacts of natural and human induced stresses upon the system. Later on, a transient model on the same aquifer was developed by (Ameen, 2019) keeping the steady-state model developed by Trasviña (2018). Ameen (2019) addressed the potential consequence and impact of future climatic variations on groundwater development, and assessed the possibility of enhancing aquifer recharge using managed aquifer recharge (MAR). In the study, a baseline period of 9 years (2010-2018) with a monthly time step was taken. The transient model showed an overall decrease in the water levels of the southern portion of the aquifer with more visibility in the phreatic aquifer, whereas there was significant discharge in the centre and north of both aquifers. No particle tracking model has been developed on the coastal aquifers of Maputo. Also there were inconsistencies in the hydraulic parameters of the different models developed, which were not analysed. Furthermore, flow path and residence time were neither compared between models, nor were analysed in detail or validated with other techniques.

Study Area Description

This chapter discusses about the basic study area information in details. It mainly comprises of five parts: the physical location, climatic conditions, hydrology, regional geological and hydrogeological settings of the study area.

3.1 General overview

The study area, Matola wetlands, is located in the Maputo Province, southern Mozambique, Figure 3.1-1. It also falls on the southern part of the greater Maputo aquifer. The topography is predominantly flat, having gentle slopes of about 0–10 degrees, except the hilly southwestern area where maximum elevation is around 230 masl. However, slopes are higher than 20 degrees near the river valleys and on the rugged western regions. Majority of the neighbouring areas of Matola River is covered by shrub-land and grassland (Nogueira, 2017).

1. Figure 3.1-1 Geographical location and elevation of the study area

3.2 Climate

Under the Köppen–Geiger climate classification, the region has a Tropical savanna climate (Aw). Maputo is the southernmost coastal province of Mozambique having the average highest temperature (26°C) in January while the lowest (20°C) occurs during July. The temperature gradually decreases towards the high inland regions (WaterAid) (NCEA, 2015) (Chairuca et al., 2018). Seasonal temperature varies due to the seasonal fluctuations of the Intertropical Convergence Zone. The seasonal variation covers a cool dry season from April to September and a hot humid season from October to March. Highest rainfall generally occurs in the north (1,000 mm/year) and lowest rainfall occurs in the southeast (500 mm/year). The yearly average precipitation is approximately 600 mm (NCEA, 2015). Moreover, rainfall varies with topographic features also, most rainfall takes place in high elevated areas and adjacent to the coast (800-1,200 mm) (NCEA, 2015). Rainfall mainly occurs during the summer season, from November to April whereas during the dry winter season (May to September) has less precipitation. In the summer period, the region receives 50-150 mm of rainfall per month. Potential evapotranspiration remains higher than average precipitation over the whole year (NCEA, 2015) (Chairuca et al., 2018). The south region faced the most significant increase in temperature (up to 1°C over past 100 years). On the contrary, average annual rainfall has also decreased between 1960 and 2006 at a rate of 3.1% per decade. Concurrently, a sea-level rose about 3 cm at Maputo coastline between 1961 and 2001 (NCEA, 2015).

3.3 Hydrology

Matola River, the main study area of this research, is a 60 km long river with perennial wetlands to its southernmost segment. It runs from the centre of the Greater Maputo region and discharges into the estuary area of Espirito Santo, southern Maputo. The northern part of the river is ephemeral in nature. High values of EC (up to 30000 µS/cm) have been observed in the river water as well as neighbouring deep wells (FIPAG, 2012; Rosário, 2016). Besides Matola River, several important water bodies also flow through the Greater Maputo region, from small perennial streams to large transboundary rivers, for instance- Incomati, Infulene, Chulavacane and Cuenga. The downstream areas of the Matola River are extensively used for pasture, subsidence farming, urbanisation and industrial development (FIPAG, 2012)

3.4 Regional geology

The study area is situated in the southern part of the Mozambique Basin constituting terrestrial and marine deposits, which deposited during different marine transgression phases from the late Palaeozoic (Muiuane, 2007). This sedimentary basin shares the western boundary with Jurassic volcanic rocks. Various Cretaceous rocks, mostlly coastal and shallow water conglomerates, marls and shales, form the bottom part of the basin's sedimentary structure (Cendón et al., 2019).

Cenozoic sediments consisting of a broad variety of limestones, clayey marls and sandstones, cover the Cretaceous rock. Sediments formed during Mesozoic and Cenozoic transgression are found at more than 200m altitude and are overlain by Quaternary sediments in most of the Great Maputo Area (Siesser and Dingle, 1981). The Tertiary alluvial, followed by Quaternary aeolian sedimentation, has laid down after the Pliocene regression (Cendón et al., 2019). The Quaternary aeolian sediments, composed of fine sand, serve as the main surface sediment of the Maputo district. The Tertiary alluvial deposits are primarily made of carbonate units (limestones and calcarenites) and are well developed along the major rivers (Nogueira, 2017; Cendón et al., 2019). Figure 3.4-1 shows the surface geological groups, whereas the dotted boundary indicates an approximated study area of this research.

Figure 3.4-2 Regional geology of the Greater Maputo, Source: (Nogueira, 2017)

3.5 Hydrogeology

The hydrogeology of the study area is strongly related to the geology of the region. There are two main aquifers in the area: an unconfined (or phreatic) aquifer and a semi-confined aquifer. The unconfined one is found within the Quaternary aeolian sand deposits. The semi-confined

one is within consolidated sands, sandstones and carbonate rocks from Tertiary period (lower Cenozoic period). These two aquifers are separated by an aquitard with irregular thickness and comprising of marls and siltstones (Nogueira, 2017; Smidt et al., 1990). Literature suggests that the phreatic aquifer has a thickness of 5-50 m, whereas the semi- confined aquifer has 50-60 m. The aquitard thickness varies between <2 m or absent, for the latter case, the system is considered as a single unit (Smidt et al., 1990), Figure 3.5-1.

Figure 3.5-1 West - East Cross-section on the southern part of Maputo City covering Matola River, Source: (WE Consult, 2013)

Groundwater levels of the greater Maputo aquifer vary from 2 to 50 m above sea level. Higher hydraulic heads on the west and lower hydraulic heads on the east (Nogueira, 2017). The conceptual model of the aquifer system depicts that despite different local flow directions near discharge areas, the groundwater flows the general hydraulic gradient from the Western inland towards the eastern coast. Alike unconfined aquifer, semi-confined groundwater also flows from west to east. Nogueira (2017) delineated an interpolated map of recharge zones (groundwater flows from the unconfined to the semi-confined aquifer) and discharge zones (groundwater flows from the semi-confined to the unconfined aquifer with leakage-up) using differences between measured phreatic and piezometric levels measured in April 2017. From this and previous studies, it is proved that Matola is gaining water from groundwater seepage. Figure 3.5-2 presents the interpolated map attained with measured water levels.

Figure 3.5-2 Groundwater head difference map between the phreatic and semi-confined aquifers. Areas in blue or positive values represent zones where groundwater flow downwards (phreactic level is higher than piezometric level); areas in red or negative values represent upward flow (phreactic level is lower than piezometric level). Source : (Nogueira, 2017).

Nogueira (2017) in his findings further presents potential groundwater recharge which ranges from 5%-15% of total annual precipitation. His study found inland high salinities near the coast and in some sections of the Western and Northern parts of the Greater Maputo aquifer, Figure 3.5-3. According to the findings of Nogueira (2017), Matola River is saline-brackish which makes the river water unusable. And this study focuses mainly on Matola River and its adjacent wetlands.

Figure 3.5-3 EC distribution of both aquifers. Blue colour = fresh groundwater; red colour = brackish. Source: (Nogueira, 2017)

Research Methodologies

This chapter describes the research methodologies and tools were used to achieve the research objective. It has four parts- data collection, hydrochemical analysis, tracer age and model residence time estimation and sensitivity analysis of the models.

To address the objectives stated in section 1.3, multiple techniques were combined in this study. Consequently, the research is divided into three major phases- Literature review, Field data collection and Result analysis. Each analysis tools and related research steps are presented in

Secondary data, related theses Literature and other published research Review Hydrochemical •Major ions Field & •Water stable isotopes (δ2H/ δ18O) Analysis 14 13 Lab •Radioactive isotopes ( C/ δ C) Work •Sensitivity analysis Numerical •Residence time and flow path Application Modelling estimation (PMPATH)

•Salinity evolution (SEAWAT) Objectives •Evaluation of salinisation conceptual model Conceptual model•Groundwater of salinisation residence time

•Saltwater origin and evolution

Figure 3.5-4 Flowchart of the research methodologies

Literature review and sample collection was also done for 36Cl tracer. But due to time constraints of lab result analysis, 36Cl outcomes could be merged with this present research work.

4.1 Data collection and sampling

4.1.1 Preliminary data collection The existing data of the study area on local hydrology, geology, hydrogeology, land use, groundwater level and quality, recharge areas and rates were acquired through relevant papers, theses and technical reports. The already developed steady-state and salt transport models were studied through theses carried out in 2018-2019. Chapter 2 and Chapter 3 represents the summarised findings of the secondary data. 4.1.2 Well inventory Field data collection and water sampling procedures in the adjacent areas of Matola River was conducted on April 2019. An attempt was taken to distribute sampling points evenly and to focus on the point of interests as well. 15 locations were visited in total, but water samples were collected from 13 locations. Out of 13 collected samples, 11 were groundwater samples, 2 were surface water samples from Matola River. Again, from 11 groundwater samples, 4 were ARA- Sul monitoring wells, 5 were regularly used private/ domestic wells, and 2 were auger-hole water, Figure 4.1-1. The sampling locations were established for shallow depth, approximately 30 m, and relatively deep depth, between 40 and 62 m below the surface.

Figure 4.1-1 Distribution of collected water samples during the field visit

Water levels and temperature in ARA-Sul monitoring wells were measured through an acoustic sounding probe water level meter. The measurement of water level was not possible in the private wells, which were tube wells as shown in Figure 4.1-2.

Figure 4.1-2 Examples of ARA-Sul monitoring well with two piezometers (left); and private/domestic well (right) Due to large well volume and lack of access to a pump, pumping out a volume three times of the standing well water before the sample collection was not possible for ARA-Sul monitoring wells. However, an effort to collect sample water from the well screens was made to improve the representativeness of collected samples. Water samples were collected from wells through a steel bailer. The sampling procedures for major ions and environmental isotopes were closely followed according to the guidelines of the International Atomic Energy Agency (IAEA) and Centre for Isotope Research (University of Groningen). 4.1.3 In situ measurements Greisinger portable digital conductivity meter and WTW pH meter were used for in situ measurement of EC, temperature and pH respectively. In situ dissolved oxygen was estimated via DO meter. Alkalinity was also measured on the field from unfiltered samples by both field titration and Hanna digital titration kit. Silica was measured by a Hanna digital silica checker. Cellulose membrane filters (0.45 µm) were used for cation-anion analysis. Cation analysis samples were collected in pre-acidified (10% HNO3) bottles of 25 ml, whereas sample collection bottles for anion analysis were not acidified. Besides, samples for water stable isotope analysis were collected in 1.5 mL glass vials without filtration. Some headspace was left while closing the caps to avoid changing pressure conditions during transport.

14 Dark coloured one-litre glass bottles were used for collecting water sample of both CDIC and δ13C. Bottles were thoroughly rinsed with the sample prior to collection. Minimal atmospheric contact was ensured during the sampling and by avoiding headspace within the samples. The samples were sterilised by adding five drops of I2-KI solution per 100 ml sample to avoid the growth of organic material inside the bottles. White coloured 250 mL HDPE bottles were used to collect the unfiltered water samples for 36Cl lab analysis. All collected samples, except water stable isotope ones, were stored at a low temperature to conserve sample characteristics.

4.1.4 Laboratory analysis 14 13 Major ions and water stable isotopes were analysed at IHE Delft laboratory while CDIC/δ C samples were sent to Centre for Isotope Research (University of Groningen) for analysis. 36Cl samples were sent to ANSTO (Austrial’s Nuclear Science and Technology Organisation) for lab analysis. Inductively coupled plasma mass spectrometry (ICP-MS) and ion chromatography (IC) was used to measure concentrations of major cations and anions, respectively. Major anion samples were diluted (10X and 100X) with distilled water to keep total heavy metals below 1000 ppb depending on the concentration of total heavy metals determined during ICP-MS in the laboratory. Samples for ICP-MS were diluted by several factors due to different calibration lines of Cl, SO4 and NO3-N. The results of the concentration of cations and anions were reported as mg/L. The δ13C signatures of waters were analysed by Isotope-ratio Mass Spectrometry (IRMS) and were reported as per mil (‰) with a precision of ±0.15‰. δ13C is the difference between a sample and standard 13C content as a fraction of VPDB standard value (Equation 1) (Mook and Plicht, 1999):

13퐶−13퐶푅푒푓 훿13퐶 = Equ. 2 13퐶푅푒푓

14 The CDIC samples were analysed by Accelerator Mass Spectrometry (AMS) and results reported from the laboratory were in per cent Modern Carbon (pMC), according to (Mook and Plicht, 1999). The results were denormalised against the δ13C of the graphite with an average error (±1σ) of ±0.25 pMC. The lab analysis report of 36Cl did not reach in time and therefore could not be analysed for this work. 4.2 Hydrochemical analysis

PhreeqC- Version 3 (Parkhurst and Appelo, 2013) was used to calculate ion balance errors, molalities, mineral saturation indices and pCO2. The spreadsheet “ChemDiagnostics” (Foppen, 2016) was used to classify water types which applies the Stuyfzand (1989) hydrochemical facies concept. According to Stufzand classification, water types are expressed as a single unit comprising of main type, type, sub-type and class. Moreover, four criteria, e.g. chloride concentration, alkalinity, major ions and a Base Exchange index (BEX) helped to determine the subcomponents of water type. Each subcomponent, again, has sub-levels with definite code. After the assembly, the name of the final water type appeared in the spreadsheet. Furthermore, data was plotted for visual interpretation in Piper plot, Stiff diagrams and scatter plots of major ions. Appelo and Postma (2005) recommended standard composition of seawater served as the reference for comparison of typical ion ratios. The endmembers for the creation of hypothetical mixing ratios/ranges with fresh water was taken from the previous study, (Nogueira, 2017). All the ion scatter plots, and ratio plots were constructed and compared, keeping the study of Nogueira (2017) as base plots.

4.3 Radioactive Isotope analysis

4.3.1 Hydrochemical Model 14 Each model uses different parameters to estimate C activity (Ao). Determination of CO2 and 13 HCO3- molar fractions, total inorganic carbon (TIC) and δ C were necessary parameters to apply corrections for the initial activity of 14C. For instance, Pearson model deals with the groundwater evolve chemically and isotopically by the dissolution of carbonate mineral under closed conditions only. This isotope mixing model calculates the dissolution of active carbon, i.e. the initial activity of the total dissolved carbon based on the 13C content of each species:

훿푇−훿푐 퐴표 = (퐴푔 − 퐴푐) + 퐴푐 Equation 3 훿푔− 훿푐 where δT, δg and δc are the stable isotope compositions of the total dissolved carbon, the soil 14 CO2 and the soild carbonate respectively. Ag and Ac are the initial C activity of soil CO2 and solid carbonate respectively. Used six models for this study were: Tamers' (1975) model, Mook’s (1972) model, Pearson's (1967) model, Fontes and Garnier's (1979) model, IAEA, (2013) model and Han and Plummer's (2013) model. The mathematical forms and comparison among the models described in Fontes and Garnier, (1979) and Han and Plummer, (2016) were mainly used to calculate the initial 14C activity. The formulas.for each model can also be found in the respective work. Groundwater can experience both open and closed system conditions while penetrating from recharge area to the deep aquifer. Most of these models assume that DIC in groundwater is derived mainly from the dissolution of soil CO2 (from root respiration) in the unsaturated zone. Later, along the flow path, the subsequent dissolution of carbonates or CaCO3/CO2 equilibrium with the aquifer matrix (closed system) is the predominant process to impact δ13C values and 14C in the aquifers (Geyh, 2000). In general, along the flow path δ13C values get enriched with a consequent decrease in 14C values (Geyh, 2000). However, Oszewski and Zuber, (1991) pointed that if the isotopic exchange partially takes place on the previously modified solid surface, the exchange process does not necessarily change the δ13C value of the DIC. As a result, the 14C can take longer time to travel without notable changes in the δ13C values. In this case, the overestimation of 14C age is possible.

Table 4.3-1 Presumed parameter values used in the hydrochemical models

Parameter Typical literature values The value used in this study 14 C activity of the soil CO2 97-100 pMC 100 pMC

14C activity of the solid 0-3 pMC 0 pMC carbonates

13 δ C content of soil CO2 -10 to -25‰ VPDB -15‰ VPDB

δ13C content of solid carbonates -10.8 to -3.2‰ VPDB -4.7‰ VPDB

14 The values of C activity of the soil CO2 and solid carbonates of this study were taken similar 13 to most of the works of literature. δ C content of soil CO2 was taken as -15‰ VPDB according to the present-day vegetation of the study area. Conferring by (Fontes and Garnier, 1979; Han, Plummer, 2012), vegetation of tropical regions (C4 native grass) mostly follow Hatch-Slack photosynthesis cycle and have an average δ13C value close to -15‰. δ13C content of solid carbonates were taken from (Peché, 2012). He reports δ13C values of Quaternary sand dunes ranging from −10.8 to –3.2‰ and an average of −4.7‰ for the sedimentary materials of Inhaca Island, near the study area.

4.3.2 Graphical method Six models resulted in a wide range of estimated groundwater age of the samples. Therefore, the graphical analysis method of Han et al. (2012), along with chemical analysis, was adapted to identify the appropriate correction for each of the samples. This study follows the steps in selecting a suitable model for the groundwater system described in Han, et al., (2012); Han and Plummer, (2016). The summarised steps are- 1) The key 14C and δ13C values for the system, for instance carbon isotopic values of the soil gas CO2, the carbonate rocks, temperature, pH, etc. were determined from literature and field data 2) Values of 14Cand δ13C for points A1 (with Eq. (4) and (5)), A2 (with Eq. (6) and (7)), O (with Eq. (8) and (9)) and M” (with Eq. (10) and (11)) were calculated 3) The points calculated in second step were plotted on the graph as described in (Han, Plummer, 2012). 4) The samples were plotted in the Han and Plummer diagram using determined values of 14 13 C, δ C and DIC (or HCO3−) concentration. 14 13 5) The data ( CDIC, δ CDIC and DIC/HCO3−) were analysed using the Han & Plummer base diagram 6) Referring to the base diagram, decision was taken which hycdrochemical model is best suited for each sample. 7) When none of the single-sample-based models could be used, two facts were considered (i) the water sample is a mixture of distinctly different aged waters, or (ii) more complex processes are taking place in the system than indexed in (Han and Plummer, 2016) 13 13 훿 퐶푎1 = 훿 퐶푔 + Ɛ푎 Equation 4 푔

14퐶푎1 = 14퐶푔 + 0.2Ɛ푎 Equation 5 푔

13 13 훿 퐶푎2 = 훿 퐶푔 − Ɛ푔 Equation 6 푏

14퐶푎2 = 14퐶푔 − 0.2Ɛ푔 Equation 7 푏

14퐶푖 ≈ 0.5 ∗ 14퐶푔 ≈ 0.5 ∗ 14퐶푎1 Equation 8

13 13 13 훿 퐶푖 = 0.5(훿 퐶푎1 + 훿 퐶푠) Equation 9

13 퐶푎 13 퐶푏 13 훿 퐶0 = ( ) (훿 퐶푠 − Ɛ푠 ) + ( ) (훿 퐶푠 − Ɛ푠 ) Equation 10 퐶푇 푎 퐶푇 푏

퐶푎 13 퐶푏 14퐶0 = ( ) (훿 퐶푠 − 0.2Ɛ푠 ) + ( ) (14퐶푠 − 0.2Ɛ푠 ) Equation 11 퐶푇 푎 퐶푇 푏 The fractionation factor values (Ɛ) were taken from (Han and Plummer, 2016). Please see Annex II for the table of parameter description.

4.4 Groundwater Flow Model Development

The purpose of the groundwater flow model study is to simulate groundwater transit times and corresponding chloride concentrations in Matola River and its adjacent wetlands. A particle- tracking method (PMPATH) was used to identify the source areas and to calculate residence times of the groundwater that discharges to Matola wetlands and individual transects across the stream by backward tracking. This modelled groundwater residence time was to be compared to that of 14C estimated residence time. The overall steps are listed in Figure 4.4-1.

Steady state Sensitivity model re- analysis of adjustment PMPATH

Model vs tracer 14C age residence time calculation

Evaluation of steady-state model

Figure 4.4-1 The overall steps followed to determine the residence time of groundwater Similarly, the transit time of variable-density groundwater flow (with chloride concentration) simulated from solute transport (SEAWAT) model was to be compared with the MODFLOW steady-state model simulated transit time. Processing MODFLOW 8 (PM 8), developed by Simcore software which provides a user-friendly graphical interface was used for the calibration of MODFLOW and SEAWAT models (Simcore Software, 2012). The steps are shown in Figure 4.4-2.

Salt transport Concentration Sensitivity model difference Run

Evaluate the hypothesis of “Paleo” fossil SEAWAT seawater entrapment SEAWAT

“Modern” Predict future SEAWAT condition

Figure 4.4-2 The overall steps followed to determine the origin and evolution of saltwater

4.4.1 Evaluation of existing steady-state flow model A steady-state model of groundwater flow in the coastal Maputo aquifer was developed by (Trasviña, 2018) based on the conceptual model analysed by Nogueira (2017). In the same research, he presented a solute (chloride) transport model using SEAWAT. The steady-state MODFLOW model calibration was done with groundwater head observations and drain discharges using manual calibration and PEST analysis. This model was further improved and used as the basis for a transient MT3DMS model developed by Ameen (2019) which had monthly inputs of precipitation data and recharge estimation of the predefined twelve recharge zones. This transient model was also calibrated to observed groundwater head. The steady-state model configuration was kept as same as (Ameen, 2019). The details description of the basic steady-state model configuration has given in Annex I. 4.4.1.1 Model Re-adjustment The age of a groundwater seepage gaining stream increases from banks to its centre. The age also becomes successively older with distance from upstream to downstream (Modica et al., 1998), Figure 4.4-3. This general assumption was applied to this study to evaluate the validity of this assumption on Matola wetland, which is a drain in the model.

(a)Conceptual model of the distribution of the flow paths, (b) Age distribution of groundwater seepage along transects of the hydraulic head, and age of groundwater seepage surrounding upper, middle and lower reaches of a stream. a stream channel Figure 4.4-3 The distribution of flow paths and age of groundwater seepage into a stream, adopted from Modica et al. (1998) The elevation of the Matola drain had some inconsistency due to the approximation in the alignment process in PMWIN. Those inconsistent elevations were corrected concerning the adjacent cell elevations to the drain. As no data were available for the depth of the Matola riverbed, in general, 2m difference was kept between the drain elevation point and the surrounding top of layer elevation points. It was also ensured that both the drain elevation and top of layer elevation decreased from upstream to downstream of the Matola River.

4.4.1.2 Model Sensitivity A manual sensitivity analysis of the model regarding the residence time of groundwater was carried out on the steady-state model to analyse its responses to model parameters like- horizontal hydraulic conductivities and effective porosity. These parameters are very sensitive

in the relative groundwater age distribution (Tóth, 1963). Therefore, hydraulic conductivity and effective porosity were changed to observe the model sensitivity with respect to groundwater residence time. Four simulations of the parameter values were used from (Trasviña, 2018) and Ameen (2019) calibrated steady-state models. Since the irregularities of the width of the layer 2 (aquitard) can impact more on the groundwater residence time, a sensitivity analysis was more focused on this layer, shown in Table 4.4-1. In all simulations, vertical hydraulic conductivities were defined as one-tenth of horizontal hydraulic conductivity values.

Table 4.4-1 Simulation of different hydraulic parameter values of the aquitards in different sectors of the aquifer for manual sensitivity analysis

Zones Horizontal Hydraulic Conductivities Initial (Ameen, Simulation 1 Simulation 2 Simulation 3 Simulation 4 2019) Young coastal dunes 0.15 0.15 1 1 0.15 Incomati River valley 0.12 0.12 0.1 0.1 0.12 Old sand dunes 0.17 0.17 1 1 0.17 Northern area 0.15 0.15 1 1 0.15 Matola River 0.06 0.06 1 1 0.06 Western zone 0.0015 0.0015 0.5 0.5 0.0015 Effective porosity Initial (Ameen, Simulation 1 Simulation 2 Simulation 3 Simulation 4 2019) Layer 2 0.25 0.15 0.15 0.25 0.15 (Layer 4) (0.25) (0.25) (0.25) (0.25) (0.15)

4.4.2 Particle tracking flow path analysis To achieve one of the objectives of this work, a steady-state particle-tracking MODFLOW- PMPATH model was developed. The improvised MODFLOW model of (Ameen, 2019) was re-adjusted, as described in 4.4.1.1. Next, transit time distributions (TTDs) were generated for nine locations from where water samples were collected for 14C analysis, including piezometer wells, private wells and Matola river water. The advective transport model PMPATH was used to retrieve the groundwater models and simulation result from PM and MODFLOW. a semi- analytical particle-tracking scheme was used by PMPATH to calculate the groundwater flow paths and residence times (Pollock, 1989). Cells with corresponding coordinates were determined first in the phreatic and semiconfined aquifers. For phreatic aquifer, 16 (4 x 4) particles were set on three faces- face 1, face 2 and face 6 of each of the cell. Similarly, for the semi-confined aquifer, 96 (4 x 4) particles were set on all the six faces of each of the cell. Different colours were assigned to particles passing through different layers to differentiate different flow paths, Figure 4.4-4. Both forward and backward particle tracking were executed for steady-state and SEAWAT model simulations. The retardation factor of the particles was specified as 1 which was used to adjust the average pore velocity of the groundwater flow.

a) b)

c) d) Figure 4.4-4 PMPATH settings; a) 6 faces of a cell; b) 48 particles in a cell of phreatic aquifer; c) 96 particles in a cell of semi-confined aquifer; d) colours provided in different layers, where blue represents phreatic aquifer, orange represents aquitard, cyan represents semi-confined aquifer and yellow represents bottom aquitard.

Each particle tracking step had a time length of 50 years. The particle tracking simulations has proceeded until all particles reached the end (forward tracking) or the onset (backward tracking) of the specified maximum time steps. All particles left from the cells via sinks were ensured. Lastly, the simulated transit time of each location was determined by taking the mean value of the transit times of predefined 48 or 96 particles.

4.4.3 Salinisation Evolution Model (SEAWAT) 4.4.3.1 Model Configuration Two steady-state salt transport models (SEAWAT) were developed from the calibrated steady- state model of Ameen (2019) to evaluate the saltwater entrapment hypothesis. Simulation settings on both of the model was “Variable density flow and Transport with SEAWAT” with no kinetic reaction. Chloride was the simulated species and the value of fluid density to solute concentration slope (DRHODC) was 0.00132. The advection package was 3rd-order TVD (ULTIMATE) solution scheme. The initial longitudinal-vertical and horizontal-transversal dispersivities defined in the dispersion package of the models are shown in Table 4.4-2. Table 4.4-3 represents the effective porosity values used in the model.

Table 4.4-2 Initial dispersion values

Layer αL (m) αTH (m) αTV (m) 1 10 1 0.1 2 10 1 0.1 3 10 1 0.1 4 1 0.1 0.01

Table 4.4-3 Effective porosity values of different layers

Layer Effective porosity 1 0.25 2 0.15 3 0.25 4 0.15

4.4.3.2 Source and Sink Concentration The sources and sinks of the chloride concentration in the model were taken from (Trasviña, 2018). According to (Trasviña, 2018)- . General-Head boundary and constant head cells represent Indian ocean and have a concentration of 19,000 mg/l. . The Incomati river chloride concentrations were defined from the previous study. The input chloride concentration values can be found in (Trasviña, 2018). . The chloride recharge concentration from precipitation was acquired through the empirical equation where the concentration was a function of distance from the ocean. Recharge chloride concentration for each zone is listed in (Trasviña, 2018). . Simulations of the potential connate saline water trapped in different locations of different layers were examined and also used for initial concentration sources. The locations in different layers having entrapped hypothetical seawater are- - The upper aquitard and semi-confined aquifer (layer 2 and 3): western sector of Matola River and northern boundary adjacent portions - Bottom aquitard (semi-permeable) layer (layer 4): fully-saturated

4.4.3.3 Initial concentration The only difference between two SEAWAT models was the initial concentration of the locations. The aim of the paleo SEAWAT model was to replicate the aquifer salinity distribution after the Holocene marine transgressions (hypothetically), where the fossil saline water are thought to be entrapped. While the modern SEWAT model was developed with present-day salinity distribution to examine the future evolution of aquifer salinisation. The initial chloride concentration for all the layers of the two models is shown in Annex III. . Modern SEAWAT model: Trasviña (2018) found that with the given the sinks and sources, after 100 years, the concentration at the observations reached a relatively steady condition. Therefore, this 100 years simulated preliminary natural model concentration was taken as the initial concentration of the present SEAWAT model, which was run for 5000 years as a future scenario.

. Paleo SEAWAT model: The initial concentration of Western sector of Matola, Northern sector of upper aquitard and semi-confined units (Layer 2 and Layer 3 respectively) and the whole semi-permeable unit (Layer 4) was given 19,000 mg/L as model initial

concentration input. This model was named as paleo SEAWAT model and was run for 5000 years to evaluate the hypothesis of entrapped paleo saline water.

4.4.3.4 Sensitivity Run Before pursuing an approximation of probable year which can be compared with present salinity distribution in phreatic and semi-confined aquifers, an effort was put to check the sensitivity of the SEAWAT model concerning effective porosity. This was done to improve the investigation of the desired results. Concentration-time plots were initiated on the locations where the chloride concentrations are known. Two simulations of effective porosity were simulated to examine the sensitivity of the model. Since the SEAWAT model was based on the calibrated steady-state model, the values of effective porosity were not given more than the calibrated values in the sensitivity analysis. Table 4.4-4 describes the simulations of the models.

Table 4.4-4 Sensitivity analysis of SEAWAT model concerning effective porosity

Initial Simulation 1 Simulation 2 Simulation 3 Layer 2 0.15 0.25 0.15 0.25 Layer 4 0.15 0.15 0.25 0.25

Results

This chapter discusses the results in three broader aspects of groundwater salinity- hydrochemical evolution and flow models evaluation, age estimation and lastly comparison between tracer age and model-simulated residence time.

5.1 Hydrochemistry

The electrical conductivity (EC) of groundwater samples ranges widely from 1100–29,300 µS/cm, whereas the EC of surface water samples (two Matola River water samples) ranges from 20,100-25,200 µS/cm. The pH of groundwater samples ranges from 6.5-7.5 (mean value 7); the pH of surface water ranges from 7.2–7.8 (mean value 7.5). Therefore, most of the samples are slightly basic. Groundwater temperature ranges from 25.4–27.6⁰C with an average of 26.5⁰C, meanwhile surface water temperature ranges from 26-26.4⁰C. The chloride concentration is as high as 1320-8400 mg/L (7.04-250.7 mmol/L). Annexe V summarises the physiochemical parameters of all the samples.

5.1.1 Water types According to the Stuyfzand classification (1989), water samples vary from fresh to brackish- salt as the main type. All the samples have moderately high alkalinity (code 3 for type), except Auger 4, where alkalinity is very high (code 5 for type), Annexe V. The location of water samples (labelled) along with Stiff diagrams is shown in Figure 5.1-1.

Figure 5.1-1 Stiff diagrams of the samples where red polygons are piezometer wells, granular polygons are surface water, pale green polygon is auger-hole water and striped polygons are domestic (private) wells. The Stiff diagram plot reveals that the majority of the samples have saltwater composition while four semi-confined aquifer samples, PZ08, PZ15, M1 and M5, have fresh to slightly brackish water. The Stiff diagram map also shows that brackish-saline waters are located on the west side of the Matola River and the freshwater (PZ08, PZ15, M1 and M5) are located on the east side of the river. It further indicates that shallow water (phreatic aquifer and surface waters) have comparatively higher chloride and sulphate concentration than deep groundwater (semi- confined aquifer samples). The major groupings or trends of the dataset are discerned visually by Piper plot, and it indicates that water samples evolve towards Na-Cl poles. The compositions of the major ions are shown in percentages. From Figure 5.1-2, Na+ is the dominant cation, albeit (Ca2+ + Mg2+) dominates over Na+ in some samples. On the other hand, Cl- is the dominant anion in all the samples. Mean distribution of the major ions of different groundwater samples results in a sequence of anions and cations are Cl>HCO3>SO4>NO3 and Na>Mg>Ca>K, respectively.

Figure 5.1-2 Piper plot of the collected samples from Matola wetlands, the red line indicates saltwater intrusion, the blue line indicates freshening, black line indicates mixing (after Han et al., 2011).

5.1.2 Major ions trend Various hydrochemical evolution of groundwater like mixing of different waters, dissolution or precipitation of minerals and cation exchange, can be assessed by plotting ion concentrations against each other, in particular against Cl as an indicator of salinity. In this study, the twelve collected samples were plotted with previous data examined by Nogueira (2017) to check the consistency of the two studies. In his research, Nogueira (2017) categorised seventy water samples into six groups, WT-1 to WT-6, where WT-1, WT-2, WT-3 are freshwater samples and WT-4, WT-5, WT-6 are brackish/saltwater. Figure 5.1-3 (a) shows that Na has a strong linear correlation with Cl and there are additional Na in freshwater. Groundwater samples plotted near dissolution line (1:1) reflects the significant addition of halite dissolution in groundwater mineralisation. With increasing salinity, most of the samples of this study deviate from 1:1 line seems to be in line with marine molar ratio (Na/Cl=0.86). PZ08, PZ15 and M5, clustering with WT-3, are above marine ratio or seawater mixing line, whereas M1 are with WT-4 cluster and M2, M3 plot around WT-6. Among these, PZ15 is slightly above the standard ratio line, indicating the enrichment of Na due to cation exchange with Ca or Mg in the aquifer during freshening. Silicate weathering could be another reason for Na enrichment. PZ16 is on marine ratio line, which indicates seawater intrusion has not reached to cation exchange yet. On the other hand, PZ14, M4, Matola

water samples and Auger-hole water samples further deviate from the marine ratio line and clustering with WT-5 of the previous study. The enrichment of both Na and Cl concentration of this group indicates seawater intrusion, followed by cation exchange. While plotting the samples in Na/Cl against Cl, six samples show an increase in Na/Cl ratio and fall above or on the marine ratio line, while the majority of the brackish samples fall below the marine line, Figure 5.1-4(a). Increasing Na/Cl ratio in brackish to near fresh conditions suggests possible cation exchange due to freshening, which can also be seen in the Piper plot, Figure 5.1-2. On the other hand, decreasing Na/Cl ratio with salinity indicates cation exchange as a result of salinisation, which is the case for both surface and deep groundwater, i.e. Matola River waters, Auger 4, PZ14, M3 and M4 respectively.

In Ca-HCO3 plot, Figure 5.1-3 (b), samples like- M4, Matola 1 and 2, and PZ16 with high salinity, lie above the dissolution line with WT-5. Similarly, Ca also increases with Cl in brackish or salty water, Figure 5.1-3 (c). This excess Ca concentration is the indication of calcite dissolution or cation exchange when freshwater and saltwater mix. Some shallow fresh and brackish groundwater deviate from the dissolution line. High HCO3 concentration of these samples could be related to chemical reactions of the recharge water and soil CO2 in the shallower subsurface during penetration and calcite dissolution. This plot can further be validated by the Ca/HCO3 vs Cl plot where saltwater samples spread further from the marine line with the increase of salinity, and brackish water or less saline water samples remain under the line, Figure 5.1-4(c).

Ca and SO4 display a moderately positive correlation, but the points are located above the marine ratio line, indicating an excess of Ca, Figure 5.1-3 (d). The points also follow the gypsum dissolution line. The enrichment of both Ca and SO4 could be an indirect result of gypsum solubility. Persisting high alkalinity with increasing salinity, particularly confirmed by the new samples can be seen in Figure 5.1-3 (e) and among them, Auger 4 have exceptionally high alkalinity. It illustrates that the following salinisation there was access to CO2 along the flow path for bicarbonate to dissolve, either because it happened in the unsaturated zone or there was fossil carbon in the aquifer that could be oxidised. Figure 5.1-3 (f) describes that with the increase in salinity, the less salty water of private wells, along with PZ14 fall on the marine line showing good correlation, whereas PZ15, PZ16 and surface waters are enriched with SO4. Enrichment of sulphate can also be seen in Figure 5.1-3 (g). SO4 in salty water increases with the decrease of Fe. Therefore, the excess SO4 might be linked to gypsum dissolution or the presence of FeS2 (pyrite), and its oxidation process which produces SO4. Again. The samples lying below the marine ratio line (Auger 4) in Figure 5.1-3 (f) could be the indication of sulphate reduction in a closed system (deeper aquifer). Generally in O2 depleted deep aquifer sulphate reduction occurs after iron reduction and before methanogenesis.

Figure 5.1-3 (h) describes the relationship between pCO2 with HCO3. Generally, under open system aquifer, calcite dissolution lowers down the partial pressure of CO2, produces HCO3 as by-product and pH rises. Therefore, the points which have low pCO2 (-1.9 to -2.3) with high HCO3 is under partially open aquifer system, whereas low pCO2 associated with low HCO3 and high pH, like- M5 and PZ15 are the result of closed system dissolution.

(b) (a)

(e) (f)

(g) (h)

Figure 5.1-3 Scatter plot of major ions a) Na-Cl, b) Ca-HCO3, c) Ca-Cl, d) Ca-SO4, e) HCO3-Cl, f) SO4-Cl; g) SO4-Fe and h) pCO2-HCO3

To understand the salinity evolution processes in a better way, major ion ratio plots have also made with Cl, Figure 5.1-4.

(a) (b)

(c)

Figure 5.1-4 Scatter of major ion ratios a) Na/Cl-Cl, b) SO4/Cl-Cl, and c) Ca/HCO3-Cl

5.1.3 Saturation Indices Saturation indices within +0.5 to -0.5 range is acknowledged as in or near equilibrium due to uncertainties of chemical analysis of groundwater samples. Red dot lines in Figure 5.1-5 are the limit of uncertainties.

(a) (b)

Figure 5.1-5 Saturation indices of major minerals Saturation index in respect to calcite, Figure 5.1-5 (a), indicates all the water samples are saturated or slightly supersaturated, except PZ08, which is under-saturated. Matola River water samples are supersaturated with calcite. Saturation index for dolomite, Figure 5.1-5 (c) shows Matola River water, auger-hole water (Auger 4) and PZ16 (near the ocean) are highly supersaturated, while PZ08 is under-saturated. Rest of the samples are saturated for dolomite. Saturation indices to gypsum, Figure 5.1-5 (b) presents all samples, except Matola 1, are under- saturated and the samples located in deeper aquifer are far from the equilibrium line., which could indirectly suggest that dissolution of gypsum is an ongoing process, though seawater mixing and cation exchange could also increase the saturation index to gypsum. A similar trend is observed for halite dissolution, Figure 5.1-5 (d). However, halite and gypsum saturation indices suggest a clear trend of approaching to saturation with increasing salinities. 34

5.1.4 Water stable isotopic analysis

5.1.4.1 δ18O vs δ 2H relationship δ2H and δ18O is an effective tool to generate information about the conditions or environments of aquifers while recharging. The stable isotope (δ2H and δ18O) data for 13 collected samples is shown in Figure 5.1-6. The modern precipitation has a mean of δ2H= 8.4‰ and δ18O= - 0.86‰ and they are expressed in per mil (‰) with respect to Vienna Standard Mean Ocean Water (VSMOW). The table enlisting all the values is given in Annexe VI. The global meteoric water line (GMWL: δ2H=8*δ18O+10) are plotted together with the local meteoric water line (LMWL: δ2H = 8.7*δ18O+15.5). The rainfall data were obtained from the meteoric stations located in Maputo (n=4, R2 =0.83).

Figure 5.1-6 δ 18O vs δ 2H plot

The observed δ2H and δ18O values of the groundwater samples, except PZ08, demonstrate a narrow range of depletion: 0.6 ‰ to -22.4‰ for δ2H and 0.93‰ to -4.35‰ for δ18O. Most of the water samples mainly cluster around the LMWL but with more depleted isotopic composition, indicating a different recharge period (colder period) than present rainwater. Mixing with other sources of water can also not be neglected. On the contrary, Matola River water, Auger 4 and PZ08 are heavier than local precipitation and lie below the LMWL indicating possible evaporation effect, which was expected and observed previously (WT-5).

5.1.4.2 δ18O vs Cl relationship Stable isotope of water, δ 18O and Cl are treated as conservative parameters to identify the possible mixing processes between freshwater and seawater. Appelo and Postma (2005) constructed a mass-balance equation to define the mixing which is expressed as:

( ) 푚푖푚𝑖푥 = 푚푖푠푒푎 × 푓푠푒푎 + 푚푖푓푟푒푠ℎ × 1 − 푓푠푒푎 Equation 12 where the end member values were taken from the previous study. Figure 5.1-7 shows the δ18O vs Cl plot, in which samples are scattered around the seawater- rainwater (freshwater) purely mixing trajectories. With the substantial escalation of δ18O with enhancing Cl concentrations; Matola River water and groundwater (Auger 4) samples show enriched isotopic composition like WT-5, showing evaporation, after mixing with seawater. Matola 2 shows the most enriched isotopic composition (δ2H = +8.8‰ and δ18O = +0.93‰), suggestive of high evaporation that can further increase the salinity which is already high.

Figure 5.1-7 δ18O vs Cl plot

7 out of 13 samples fall within the freshwater-seawater mixing zone and PZ15 SC plots close to the mixing line range (1.2% mixing ratio), indicating the contribution of mixing with seawater in these areas. On the other hand, for PZ14 SC and PZ16 SC it is difficult to distinguish between halite dissolution and mixing with older seawater because of their depleted isotopic composition with increasing Cl concentration.

5.2 Groundwater flow model

5.2.1 Steady-state Model Re-adjustment After improving the elevation of the drain Matola, comparison between the initial model and the re-adjusted model was made in regards of hydraulic head in the phreatic and semi-confined aquifer, water budget, groundwater flow path and residence or transit time groundwater. 5.2.1.1 Hydraulic head The simulated hydraulic heads of the phreatic and semi-confined aquifer showed minor variations as compared to the previous model on the west side of the Matola River, which can be neglected. Figure 5.2-1 shows the hydraulic head differences in both aquifers.

Phreatic aquifer (initial) Semi-confined aquifer (initially)

Semi-confined aquifer (after calibration)

Phreatic aquifer (after calibration) Figure 5.2-1 Comparison of the hydraulic head after calibrating the model with the changed drain elevation for both phreatic and semi-confined aquifers

5.2.1.2 Water budget The re-adjusted steady-state water budget presented almost similar results as the initial model. In the re-adjusted model, recharge remained the same but discharge to sea and drains increases, whereas river leakage decreased in comparison to the initial model. Still, groundwater discharge to the rivers and drains of the basin was much higher than that of the sea. A comparison between the two models is indicated in Table 5.2-1.

Table 5.2-1 Water budget for steady-state models

Re-adjusted model Initial model Flow term (m3/day) In Out In-Out In Out In-Out Sea 0 92720 -92720 0 91221 -91221 Wells 0 0 0 0 0 0 Drains 0 458174 -458174 0 456936 -456936 Recharge 929136 0 929136 929136 0 929136 ET 0 0 0 0 0 0 River leakage 11805 265475 -253669 11799 267736 -255937 Head dependent boundary 0 124572 -124572 0 125042 -125042 Sum 940940 940940 0 940934 940934 0

5.2.1.3 Flow path and residence time As discussed in 4.4.1.1, flow path and residence times at different points of the Matola River were simulated and compared between the two models (initial and re-adjusted). Difference between flow paths was not much visible, though the residence times varied, Figure 5.2-2. (a) (b)

The maximum change was noticed on the point Matola 1, where the residence time decreased by 24%. The comparison of residence times between the two models is shown in Table 5.2-2.

Table 5.2-2 Comparison of residence times after re-adjustment (along distance from upstream)

Residence Time (Years) Distance Point name (Row, Column) Initial (Ameen, 2019) After Re-adjustment 0 1 (114, 90) 115.7 110.8 11 2 (135, 81) 243.0 247.1 20.1 3 (152, 77) 589.5 621.9 30.4 4 (171, 69) 852.3 798.7 34.8 Matola 1 (180, 67) 4103.0 3143.5 41.4 5 (193, 65) 453.5 460.0 46.2 6 (202, 62) 446.4 459.0 54 7 (216, 70) 870.6 771.9 57.6 8 (228, 73) 406.4 272.5

Moreover, in both models point 1-4 showed the expected rising trend of groundwater residence time, but the rest of the points showed inconsistent results. Therefore, it can be excluded that in either model groundwater transit time of Matola wetlands becomes progressively higher with distance downstream as expected, Figure 5.2-3.

(a) Initial residence time (b) Residence time after re-adjustment Figure 5.2-3 Residence times at different distances from upstream to downstream point of the Matola River

5.2.2 Steady-state Model Sensitivity Analysis Since no noticeable impact was found on hydraulic heads and residence times of groundwater between initial and re-adjusted models, the manual “trial-and-error” sensitivity step for residence time was carried on the initial steady-state model. The model was run with four simulations or combinations of the steady-state model values from Trasviña (2018) and Ameen (2019), as discussed in 4.4.1.2. The simulation settings are represented in Table 4.4-1. From the sensitivity analysis results, shown in Table 5.2-3, it can be seen that the model is more sensitive towards the effective porosity of the aquitard layers than that of hydraulic conductivities.

Table 5.2-3 Summary of the manual sensitivity analysis of the steady-state model

Residence Time (Years) Locations Initial model Simulation 1 Simulation 2 Simulation 3 Simulation 4 PZ08 SC 952 705 698 961 777 PZ14 SC 736 501 544 805 644 PZ15 SC 6334 6126 7654 7871 5909 PZ16 F 860 713 713 860 840 Matola 1 3144 2990 2795 2981 7444 M1 319 241 243 322 311 M2 1212 848 930 1284 977 M3 1484 1164 992 1348 1583 M5 986 752 765 976 1081

The range of residence times does not vary much, except PZ15 SC and Matola 1. The highest difference between the upper and lower limit of residence time was found on Matola 1 (surface water) location, and that is 4649 years. On the other hand, the minimum difference between upper and lower limit, 82 years, was found on M1 (private well) location. The summary of the residence time ranges is shown in Figure 5.2-4.

Figure 5.2-4 Upper and lower boundaries of the modelled residence times from the manual sensitivity analysis regarding effective porosity and hydraulic conductivities

Notwithstanding, the transit time of most of the locations was barely sensitive to effective porosity and hydraulic conductivities, it was undoubtedly sensitive towards the number of particles defined for the analogous cell in PMPATH. In the model, each cell (500 m * 500 m) was considered as one parcel of water incorporating 48 (phreatic and surface water) or 96 (semi- confined aquifer) particles. Again, each particle was considered a separate water molecule having a different travel time. Thus, one water parcel contained 48 or 96 different transit times. Transit times changed with the number of particles within their respective range at a great extent

and the distribution of travel times of each water particle was highly skewed. Therefore, the mean transit times taken as the simulated transit time of the locations were not so representative. The histogram of all the water parcels (cells in the model) are shown in Annexe VII. Instead of mean; median transit time was more representative as simulated transit time. The statistical descriptions are demonstrated in Table 5.2-4, where “n” means the number of particles set for that cell.

Table 5.2-4 Summary of descriptive statistics of the simulated mean residence times (in years) at different observation points

Location n Maximum Minimum Range Mean Median Standard deviation

PZ08 SC 96 2064 341 1723 952 711 575 PZ14 SC 96 1709 285 1425 736 549 451 PZ15 SC 96 47605 619 46986 6334 1272 11835 PZ16 F 48 9887 19 9868 860 684 1557 Matola 1 48 28378 33 28345 3144 915 7126 M1 96 704 155 549 319 232 193 M2 96 2153 933 1220 1212 1089 312 M3 96 2952 518 2434 1484 1318 681 M5 96 2143 685 1458 986 960 259

From the position of mean and median transit time, shown in Figure 5.2-5, it can be seen that the simulated mean values for PZ15 SC and Matola 1 have the highest difference with the respective median values. And mean transit time is not representing most of the transit times for these two locations. Therefore, the median values were regarded as modelled groundwater residence times for the ease of understanding. The residence times of groundwater remaining within the respective range were also considered to be acceptable, as shown in Table 5.2-4.

)

Years

time ( time

Residence

Figure 5.2-5 Upper and lower boundaries of the modelled residence times from the manual sensitivity analysis regarding particle numbers Minor changes in the flow paths of the observation points according to the simulation parameters were visualised. But the main course of way remained similar in all the simulations.

Except for PZ16, all the locations were getting both local and regional flows. More long regional flows were visible in the midstream of the Matola catchment, while the downstream had mostly short local flows. Annexe VIII shows the overall flow path trends of the observation points.

5.2.3 SEAWAT Model Simulations 5.2.3.1 Paleo SEAWAT Simulation The purpose of the paleo SEAWAT model was to have the most probable justification for the present-day groundwater salinity distribution. Sensitivity analysis gave an idea about the ranges up to which groundwater transit times might vary considering the chloride transport with groundwater. The response of effective porosity change was robust. With the initial effective porosity in both aquitard layers (0.15), groundwater salinity concentration of all the observation points reached a relatively steady condition earliest, while increasing the effective porosity only in the lower aquitard or in both aquitards showed almost similar result and arrived at the latest time. Increasing the effective porosity of upper aquitard befell the transit time in between the extreme two situations and took slightly more time than the initial paleo SEAWAT model to be stable. The difference of transit times between the extreme two conditions were approximately 1000-1500 years. Figure 5.2-6 summarises the concentration-time curves of selected observation points, simulating for 5000 years. Table 5.2-5 and Figure 5.2-8 show the overall qualitative results of the model run.

Table 5.2-5 Qualitative summary of the Paleo SEAWAT model simulation

Observation Points Result compared to present-day distribution PZ08 SC Higher PZ14 SC ≈ PZ15 SC Higher M1 ≈ M2 ≈ M3 ≈ M5 Higher PZ16 F Lower Matola 1 Lower

(mg/L)

Concentration

Time (Years)

Figure 5.2-6 Concentration-time curve of different observation points over 5000 years

Figure 5.2-7 Paleo SEAWAT model simulation results of the phreatic aquifer with different effective porosity (ne) in layer 2 and layer 4 44

Figure 5.2-8 Paleo SEAWAT model simulation results of the semi-confined aquifer with different effective porosity (ne) in layer 2 and layer

For both phreatic and semi-confined aquifers, the initial paleo SEAWAT model (with aquitard effective porosity = 0.15) showed quicker flushing of saltwater than the model where effective porosity of aquitard layers was 0.25. Ultimately, the transit time of groundwater had become shorter as well for the initial paleo SEAWAT model. The result of the initial paleo SEAWAT model was therefore too fresh as to present saltwater distribution in the aquifers. Nonetheless, when the effective porosity is increased to 0.25 for both aquitards, the model resulted in almost present-day salinity distribution after 3000-4000 years forward run, Figure 5.2-7 and Figure 5.2-8. However, the chloride concentration of phreatic aquifer (PZ16 F) and surface water (Matola 1) was quite lower than present-day distribution at 3000-4000 years. These two observation points gained concentration like present-day within 300-600 years.

5.2.3.2 Modern SEAWAT Simulation

The modern SEAWAT model with present-day chloride concentration as input was run to assess future salinity distribution around Matola River. The concentration-time curves of observation points indicated a declining trend of chloride concentration, except PZ08, M3 and M5. These three locations showed an increasing trend. Moreover, chloride concentration at M3 and M5 got stabilised nearly after 2000 years, whereas, chloride concentration at PZ08 seemed to escalate even at 5000 years. For the rest of the observations, chloride concentration initially increased up to 1000 years roughly; subsequently, concentration curves dropped gradually. Figure 5.2-9 showed the concentration-time curves for 5000 years. The chloride concentration was higher in all the observation points when the effective porosity of the aquitard layers was increased from 0.15 to 0.25.

In general, the model exhibited the freshening of the aquifers around Matola wetlands nearly after 1000 years from the current time, Figure 5.2-10. The model also showed that by the end of 5000 years, the upper aquitard and semi-confined aquifer (Layer 2 and Layer 3) would be almost free of salinity. The bottom layer would still have some salinity on the western hilly and northern part of Greater Maputo region, though the magnitude and areal extent will be much less. Nevertheless, the phreatic aquifer would confront a little more salinity than contemporary distribution after 5000 years,

(mg/L)

Concentration

Time (Years) Figure 5.2-9 Concentration-time curve of different observation points over 5000 years

Figure 5.2-10 Present SEAWAT model results of the phreatic aquifer (top row) and the semi-confined aquifer (bottom row) with different effective porosity (ne) of layer 2 and layer 4

5.3 Radiocarbon isotopic analysis (14C age estimation)

Table 5.3-1 presents the values for the δ13C measured concerning VPDB for each of the water samples. δ13C data for water samples are relatively consistent and range from –11.2 to –14.9‰; except for PZ08 SC, which has an enriched δ13C value of -2.04‰. The corresponding pMC estimate of 99.90% is the second-highest of all the studied locations. The pMC values of the 14C show a wide range, between 46.77 to 105.42% for the rest of the samples. Two shallower water,

Table 5.3-1 Isotopic data of the sampling sites;

- Number in pH Temp HCO3 CO2_aq TDIC δ13C ± 1σ 14C ± 1σ Figure 5.3-2 (K) (mmol/kg) (mmol/kg) (mmol/kg) (‰) (pMC) PZ08 SC 1 6.5 300.8 2.5 1.5 4.0 -2.04 ± 0.15 99.90 ± 0.30 PZ14 SC 2 6.9 299.4 6.1 1.0 7.2 -13.50 ± 0.15 70.15 ± 0.24 PZ15 SC 3 7.5 297.6 5.3 0.3 5.6 -14.93 ± 0.15 57.31 ± 0.20 PZ16 F 4 7.2 299.8 7.4 0.5 7.8 -11.47 ± 0.15 79.78 ± 0.24 Matola 1 5 7.2 299.2 5.0 0.3 5.3 -11.24 ± 0.15 105.42 ± 0.30 M1 6 7.2 299.8 6.5 0.7 7.2 -14.39 ± 0.15 77.75 ± 0.24 M2 7 7.1 299.1 6.0 0.8 6.9 -14.26 ± 0.15 76.14 ± 0.24 M3 8 7.1 297.8 6.9 0.8 7.7 -14.53 ± 0.15 46.77 ± 0.20 M5 9 7.3 298.6 4.3 0.4 4.7 -12.81 ± 0.15 74.34 ± 0.27

The isotopic data of Table 5.3-1 and the previously discussed parameter value of Table 4.3-1 were input to the hydrochemical models to estimate the initial activity of 14C (Ao), showed in Table 5.3-2. Model output results in terms of apparent ages are presented in Table 5.3-3. For ease of comparison among the overall performance of the hydrochemical models, a mean initial activity value for each of the models was calculated. Figure 5.3-1 graphically presents these results.

Table 5.3-2 Calculated initial activity (as pMC) of water samples, using six hydrochemical models.

Observation Number Tamers Mook Pearson F&G IAEA Han & Plummer Points in Figure 5.3 2 PZ08 SC 1 69 95 -26 69 28 -10 PZ14 SC 2 57 69 85 57 190 108 PZ15 SC 3 52 56 99 52 216 124 PZ16 F 4 53 78 66 53 161 92 Matola 1 5 52 103 64 52 159 90 M1 6 54 76 94 54 202 118 M2 7 56 75 92 56 202 115 M3 8 55 46 95 55 210 118 M5 9 54 73 79 54 183 104

The Tamers, Mook, Pearson and Fontes-Garnier models had the initial activity of 14C in between 54-99 pMC whereas IAEA model demonstrated to grossly over-predict the initial activity. Han and Plummer model also showed initial 14C activity more than 100 pMC. Since 49

14 the initial C activity of the soil CO2 gas is assumed to be 100 pMC, the excess pMC values (more than 100 pMC) illustrate solid-phase exchange under saturated zone dominating systems.

IAEA

F & F

Mook

Tamer

Han & & Han Plummer Pearson Figure 5.3-1 Average initial activity (as pMC) by each model for water samples

14 14 Different models resulted in a wide range of the initial C activity ( C0). Each model considers various processes that could have affected the 14C data as described in section 2.2.2. To determine the possible reactions and the best approach for each model, the graphical analysis method was applied based on the chemical and isotopic data from the studied groundwater samples. The graph of Figure 5.3-2 displays a mixing effect by plotting δ13C vs 14C content. The graph is called a Han and Plummer plot, and the summarised steps for plotting the graph are described in section 4.3.2.

(a)

(b) Figure 5.3-2. a) Graphical presentation of data from Matola catchment. Sample numbers are according to Table 5.3-1; b) Han and Plummer plot from Han and Plummer, (2016).

In Figure 5.3-2, A and A1 represent CO2 (g) and CO2 (aq) respectively. Again, A2 represents points where only HCO3 of sample is in equilibrium with CO2 (g). Again, A3 represents the stage where both CO2 (aq) and HCO3 are in equilibrium with CO2 (g). The point M serves as solid carbonate. Therefore, M′ represents only HCO3 is in balance with solid carbonate. And the point M″ denotes a mixed sample where both CO2 (aq) and HCO3 of waters are in balance with solid carbonate. Point “O” is denoted as Tamers point which represents the initial carbon isotopic composition of the system prior to carbon exchange with solid carbonate. A, O and M points are the end members in the groundwater system. The 14C activity of soil CO2 was considered as 100 pMC in this study. Therefore, the area above 100 pMC line is considered as open system (unsaturated zone) and so water samples incorporates only CO2 (aq) 14 equilibrated with soil CO2 at the point “A”. The area below the 100 pMC of C is considered as the beginning of the closed system or saturated zone. At the intersection point of X-Y (“O”), the infiltrated water has reacted totally with soil CO2 and partially with carbonates in a closed system but the isotopic exchange have not taken place yet. Thus, water samples plotting near point O are often comparatively “juvenile” water in the closed system. At “M” point, the DIC 51

in water has enriched δ13C or very low 14C compared to that of point “O” because of complete reaction with solid carbonates and carbon exchange. This plot can also be verified by chemical data of the samples Annexe V and Figure 5.1-3 (h). Considering Han and Plummer (2016), all the samples plot under the Pearson area, except PZ08 SC and Matola 1, which are number 1 and 5 respectively on Figure 5.3-2(a). Therefore, Pearson proposed correction would not be applicable for these samples. PZ08 SC (“1” in Figure 5.3-2.a) plots above the zero-age area. Similarly, Matola 1 (“5” in Figure 5.3-2.a), plotting above 100 pMC 14C line is considered as contaminated water. From Table 2.2-1 (Appendix 5-A), it can be seen that all the samples but PZ08 SC are in equilibrium with calcite saturation. The low partial - pressure of CO2 (g), high HCO3 with low pH indicates a partially open groundwater system for PZ08 which validates the possibility of a mixture with a fraction of 14C-bearing infiltration - water. One the contrary, low partial pressure of CO2 (g), low HCO3 along with high pH indicates the groundwater systems in closed conditions. Discussion about the open-close groundwater system can be found in Figure 5.1-3(h). Considering the geochemistry and different models assumptions, Pearson model and Han- Plummer model were considered for apparent ages of the sample waters. The negative age values of the water samples in Table 5.3-3 indicates either modern water or mixtures of closed- system waters with open-system water. Underestimation of the initial 14C content might arise from this mixing, and ultimately, the age values are overcorrected.

Table 5.3-3 Ages of water samples (in years BP) calculated using six hydrochemical models

Number Tamers Mook Pearson F&G IAEA Han & in Figure Plummer 5.3-2 PZ08 SC 1 -2994 382 --- -274 -10178 --- PZ14 SC 2 -1651 3036 1584 142 8011 3459 PZ15 SC 3 -688 4637 4418 1346 10664 6183 PZ16 F 4 -3299 2036 -1556 -1570 5621 1148 Matola 1 5 -5620 -199 -4074 -3515 3290 -1264 M1 6 -2856 2196 1532 -1055 7650 3385 M2 7 -2479 2366 1591 -647 7829 3370 M3 8 1337 6276 5730 3065 12054 7448 M5 9 -2607 2582 462 -411 7220 2666

Discussion

This chapter tries to give proper explanations of the results obtained which in turn will help to address the research questions. It discusses 1) the conceptualisation of prominent processes controlling the groundwater chemistry, 2) the performance of the existing numerical models, 3) the comparison between 14C age and model residence time, and 4) the origin of salinity

6.1 Conceptualisation and steady-state model of salinisation

The hydrogeological conceptual model of the greater Maputo region developed and improved by (Nogueira, 2017) and (Trasviña, 2018) was used to understand the hydrogeochemical evolution of the groundwater linking with the samples collected for this study, Figure 6.1-1.

Figure 6.1-1 Conceptual model developed by (Nogueira, 2017)and improved by (Trasviña, 2018) The groundwater system consists of four relatively homogenous layers. The first layer is unconfined or phreatic aquifer, and the third layer is semi-confined aquifer. The first layer consists of Quaternary aeolian sand deposits, and the third layer consists of consolidated sands, sandstones and carbonate rock. Both aquifers are connected through the second layer which is silty marl and clay dominated aquitard. In some parts, the second layer is either very thin or absent. In the latter case, both aquifers were treated as a single entity (Nogueira, 2017). The fourth layer at the bottom is also an aquitard. According to (Nogueira, 2017; Cendón, 2019),

the entrapped fossil water from the marine transgressions is mainly responsible for the salinity and brackishness in both aquitard layers and will be discussed later in 6.4. In the main recharge zones, rainwater infiltrates, mixes with the saline water in the bottom of the phreatic aquifer and creates a mixing zone of brackish/saline water. The bottom layer is another primary source of salinity. Being fully saturated, the salinity concentration of the fourth layer fluctuates and reduces towards the Indian Ocean. According to the conceptual model, rainwater is the primary source of groundwater recharge, and the hydrochemical evolution of groundwater starts with it. Rainwater dissolves the atmospheric CO2 (g) prior to reaching the ground and after that rainwater permeates through the soil and reacts with the CO2 (g) produced by the root respiration and oxidation of organic matter in the soil (Appelo and Postma, 2005).

퐶퐻2푂 (푂푀) + 푂2 → 퐻2푂 + 퐶푂2 (푔) Equation 13

퐶푂2(g) → CO2(aq) Equation 14

+ − 퐻2퐶푂3 ↔ 퐻 + 퐻퐶푂3 Equation 15 As the recharging water permeates through the unsaturated zone and then the phreatic aquifer, the acidic water first encounters with the aeolian Quaternary sand layers. With the presence of excess H+, silicate weathering plays a role in the solute concentration of the water. The production of CO2, shown in Equation 13, enhances the weathering of silicate minerals and increases Na+ concentration, as for instance through Equation 16. Other cations are also released. Besides these processes, the ion concentrations in the unsaturated zone increases by evapotranspiration.

+ + 2푁푎퐴푙푆푖3푂8 (푠푖푙푖푐푎푡푒) + 2퐻 + 9퐻2푂 → 퐴푙2푆푖2푂5(푂퐻)4(푐푙푎푦) + 2푁푎 + 4퐻4푆푖푂4 Equation 16

2+ − 퐶푂2 + 퐻2푂 + 퐶푎퐶푂3 (푐푎푙푐푖푡푒) ↔ 퐶푎 + 2퐻퐶푂3 Equation 17

2+ 2− 퐶푎푆푂4. 2퐻2푂 (푔푦푝푠푢푚) ↔ 퐶푎 + 푆푂4 Equation 18 Flowing further down towards the semi-confined aquifer, made up of tertiary consolidated sandstones and carbonates, water pressure and temperature increases and oxygen becomes exhausted, the dissolved oxygen values, shown in Annexe V, also validates this concept. With the presence of carbonate in the semi-confined aquifer, dissolution of calcite and gypsum take 2+ 2- place. Ca , HCO3- and SO4 concentrations increases in the water, Equation 17 and Equation 18. Later on, the water encounters the marl-clay layer (aquitard) where Ca2+ rich water from calcite dissolution might get exchanged with Mg2+ enriched clay surface, Equation 19 Furthermore, due to the presence of organic matter in some places of the anoxic aquifer, the 2- reduction of SO4 may take place, which produces H2S, observed on the field (Auger 2) by the smell of rotten eggs, Equation 20.

+ + + + 퐶푎2 + 푀푔2 − 푐푙푎푦 ↔ 퐶푎2 − 푐푙푎푦 + 푀푔2 Equation 19

2− − 2퐶퐻2푂 (푂푀) + 푆푂4 → 2퐻퐶푂3 + 퐻2푆 Equation 20 But the most crucial factor of the fourth layer is the presence of entrapped fossil seawater. Where the recharged freshwater gets in contact with saline water, mixing and freshening together act as the dominant processes. The flushing of connate saltwater by the recharge water transforms the saline system into the brackish or freshwater environment in the discharge zone

of PZ08, M1 and M5. But in other places, i.e. PZ14, PZ16, Matola River and Auger-hole samples shows salinisation still happening, as seen in the Na/Cl ratio also (Figure 5.1-4a). The trapped saltwater is indeed discharging through these areas mostly. In the discharge areas, when the groundwater comes in contact with atmospheric oxygen, it gets affected by CO2 degassing and mineral precipitation. Since the discharge of saltwater intrudes the freshwater aquifer, fed by the recharge water, areas where recharge is more significant, freshening occurs, and where saline water diffusion and discharge is prominent and recharge is limited; salinisation takes place. Mixing leads the dominant anion to be Cl, in the latter case particularly-. The concentration of Cl- additionally increases with the evapotranspiration of the discharge areas, in shallow groundwater and surface water. From the relationship between δ18O vs Cl, Figure 5.1-7, the effect of evaporation can be evaluated. Auger-hole water seems to be the discharge areas of saltwater, where the signal of direct evaporation is also very strong. Figure 6.1-1 shows the main water type of two points (Matola 1 and PZ14 SC) among 12 sampling points of this study. Linking Stuyfzand classification showed in Annexe V and Piper plot, Figure 5.1-2 with the above discussion, some important facts are evident-  The surplus of Na+ concentration in the water samples of eastside of Matola River is an indication of freshening of the aquifer along with silicate weathering. - -  The overall dominance of Cl over HCO3 indicates salinization further facilitated by the low recharge rate in the west of the study area and the effect of evapotranspiration on shallow groundwater and surface water; for instance- Matola River and Auger-hole samples.  Low SO4 concentrations, seen in PZ08 SC, Matola 2 and Auger 4, can be due to sulphate reduction from the presence of high organic matter content or microbiological activities in the soil or both. On the other hand, high SO4 concentrations, especially observed in PZ15 SC, Matola 1 and M5 can be related to mixing of seawater with freshwater in the aquifer or gypsum weathering, Figure 5.1-3 (g).  Six out of twelve samples have positive Base Exchange index (BEX) indicates freshening as the prevailing process on the locations of PZ08, PZ15, M1, M2, M4 and M5. While the negative BEX means seawater intrusion is still an effective process on West of the Matola river, Annexe V. This finding also matches with the Na/Cl ratio plot, Figure 5.1-3(a).  From the Stiff Diagram plot, Figure 5.1-1, it is also visible that the Eastside of the Matola River is fresher than the Westside. This freshening seems to be linked to the recharge rate and thickness and permeability of the aquitard as discussed below.

The existing steady-state model shows coherence with the conceptualisation of groundwater salinisation. Matola wetlands have three zones of recharge. These recharge zones were specified on the basis of land cover, studied by (Andreetta, 2018), Figure 6.1-2. The west side of Matola River has dense shrub lands which prevent the rainwater from infiltrating and therefore, has a low recharge rate (44 mm/y) on the model. On the contrary, the east side has grassland with sparse trees and henceforth have higher recharge rates on the model than the west side of Matola River. Thus the recharge flux of the model also had consistency with the hydrochemistry of the samples collected for this study. Besides recharge rate, the hydraulic conductivities of the model layers and aquitard thickness were also studied to check the validation of the numerical model regarding the freshening processes observed in the newly collected samples. Horizontal hydraulic conductivities of all layers seemed to be correlated since the east side of the river had higher hydraulic conductivities 55

in all layers than the west side, Figure 6.1-3. But the thickness of the aquitard had little impact on the samples being fresher or brackish, Figure 6.1-4.

(a) (b) Figure 6.1-2 (a) Recharge potential rates as input in the numerical model, (b) Land cover studied by (Andreetta, 2018)

Figure 6.1-3 Horizontal hydraulic conductivities as input in the numerical model 56

Figure 6.1-4 Layer thickness in the numerical model with the location of 4 newly collected samples (column view) According to Stuyfzand water type classification (Annexe V) and the Stiff Diagram plot, Figure 5.1-1, PZ15 SC and M1 are fresher samples located on the east side of the Matola River, whereas Matola 1 and M3 are saline-brackish water situated on the West side of the Matola River. Among four samples, PZ15 SC is underlain by the thickest part of aquitard, and M3 is underlain by the thinnest part. Therefore, freshening-salinisation processes cannot be adequately explained by the modelled aquifer thickness.

6.2 Residence time of groundwater

6.2.1 Model simulated residence time The steady-state model used in this study provided a two-dimensional representation of groundwater flow patterns and rates. It was developed by (Trasviña, 2018) and (Ameen, 2019) to help calculating transit times from the recharge zones to the discharge zones by the backward particle tracking method. The re-adjusted steady-state model did not show any variation in hydraulic heads, indicating no noteworthy stress on the system. So, the initial model was used for the sensitivity analysis. The overall uncertainty in model simulated travel time was defined only in a qualitative sense. The assumption of steady-state conditions does not consider transient stresses; i.e. precipitation fluctuation, the effect of drought or occasional groundwater withdrawals. The initially assigned values for porosity and recharge rate and their uniform distributions in the aquifer system all are potential sources of uncertainties. The sensitivity of the calculated travel times to changes in hydraulic conductivity and anisotropy was low. The calculation of travel time was rather sensitive to changes in effective porosity. When the possible extremes of effective porosity (lowering the value) was given as model input, the model produced a shorter travel time. Nevertheless, the model showed the most significant discrepancies to the changes in particle numbers assigned for each cell. Each cell in the model (500m * 500 m) was considered as one parcel of water incorporating several particles. Again, each particle was considered a separate water fraction having a different flow path and travel time. Thus, one water parcel had a mixture of different flow paths, from local to regional flows with different transit times and set up its own flownet. For instance, Figure 6.2-1 shows the flownet generated for Matola 1in PMPATH.

Figure 6.2-1 Flownet of Matola 1 in the model with local and regional flow paths The average travel time is the mean travel times of different water fractions and it is recognised as the weighted average value for all associated water fractions of one water parcel. It is also considered that no other process like mixing, dispersion or diffusion influences this idealised water particles along the flow path (Suckow, 2014). Similarly, in the studied model, since particle tracking took the average of all the travel times associated with one cell, increasing the number of particles henceforth achieved a more acceptable result.

6.2.2 14C calculated apparent age Tamers model considers only simple chemical mixing of two carbon atom sources under closed-system conditions without any carbon isotopic exchange processes. And this isotopic exchange might change 14C content which will not be considered in Tamers model (Han and Plummer, 2016). On the other hand, Mook model is explicitly constructed to consider only the gas phase carbon isotopic exchange processes. This model ignores the carbon exchange occurring under closed-system conditions (saturated zone) (Gallagher et al., 2000). Pearson model also counts only carbon mixing under closed conditions and neglects carbon exchange during the dissolution of CO2(g) under open conditions, alike Tamers’ model (Han, et al, 2012; IAEA, 2013). But the improvement of Pearson's model in comparison to Tamers' model is that it includes the δ13C data also. Hence, Pearson model recognises the 14C dilution by further addition of dead carbon from fossil organic matter oxidation, magmatic CO2 dissolution and cation exchange on clay minerals (Han and Plummer, 2016). Fontes and Garnier's model considers both carbon exchange and mixing under open and closed- system conditions (Fontes and Garnier, 1979). Nevertheless, Han and Plummer (2013) pointed out some disagreements in conceptualisation of Fontes and Garnier model which lead to the underestimation of 14C ages for groundwater. Instead, Han and Plummer have corrected Fontes and Garnier model by combining Tamers, Mook and Eichinger models (Han and Plummer, 58

2013). Revised Fontes and Garnier model or Han and Plummer model also acknowledges carbon mixing as well as carbon exchange existing in the unsaturated and saturated zone (Han and Plummer, 2013). However, the weakness of Han and Plummer's model is not considering the change of DIC concentration, rather considering 14C dilution caused by other processes (Han and Plummer, 2016). On the other hand, the IAEA model presumes that the isotopic exchange of the DIC occurs under closed-system conditions only after the complete isotopic exchange under open-system conditions. (IAEA, 2013). Therefore, Han & Plummer's model calculated age can be younger than the IAEA model calculated age (Gallagher et al., 2000).

14 There was a close concurrence between Co estimation of the Tamers, Mook, Pearson and F&G models, while IAEA and Han and Plummer model produced higher results. Pearson model considered only closed system, while Han and Plummer model considered both open and close system. The enriched value of initial 14C content in Han and Plummer’s model implies a significant amount of either carbon mixing or carbon isotopic exchange between the total dissolved inorganic carbon (TDIC) and the solid aquifer matrix. Hence, the residence time calculated by Han and Plummer was higher than that of the Pearson model. Notwithstanding, since vegetation cover may have changed since the time of recharge, δ13C of soil CO2 is an adjustable and sensitive parameter which has a response on the estimation of the age of the groundwater (Fontes and Garnier, 1979). Another most frequently observed and deliberate hydrochemical reaction, which disturbs the initial 14C activity of DIC is the formation of CO2 from the oxidation of organic matter in the saturated zone. The consumption of dissolved oxygen or the reduction of sulphate or the presence of nitrate in the aquifer promote the reaction. (Geyh, 2000). Once formed, the CO2 will further enhance the fossil carbonate dissolution. The fossil-derived CO2 will increase the apparent age and concentration of DIC, while the change in corresponding δ13C values will be irregular. Thus under similar situations, the measured stable isotope (δ13C) values will not be suitable either for calculating the initial 14C activity or for rectifying the secondary non-decay changes to the 14C(DIC) activity (Geyh, 2000). Figure 6.2-2 illustrates the changes of 14C and corresponding δ13C values of DIC due to fossil organic matter oxidation. The high sulphate concentration in some samples (discussed in 0) and the possibility of the presence of organic matter in some places the aquifer (for instance, Auger 4), presented in 5.1.2, and henceforward contains uncertainty.

Figure 6.2-2 Simplified plot of the changes in 14C age, DIC concentration and δ13C (DIC) values in groundwater as a result of fossil organic matter oxidation and the successive fossil carbonate dissolution. Source: Geyh (2000)

To achieve the precise estimation of the corrected radiocarbon age, geochemical mass balance approach can be applied with adequate chemical, mineralogical and isotopic data. In addition to DIC, mass balance approach takes into consideration all the geochemical reactions occurring on the flowpath of an aquifer system as well as dissolved organic carbon (DOC) and methane (CH4) (Wigley, 1975; Han and Plummer, 2016). Tracer analysis would be more reliable if the data are continuously collected for long-time period instead of a single collection. The multi- tracer approach is also necessary to cross-match and validate the apparent ages of the samples. In the beginning, this study considered two tracers- 14C and 36Cl. But later on, 36Cl tracer analysis could not be incorporated due to time constraint. Finally, the sampling technique was not very systematic due to local limitations, which puts a question on the proper representativeness of the samples.

6.2.3 Comparison of model-simulated and 14C calculated transit time The apparent ages calculated by two hydrochemical models are more correlated with the simulated median residence time than the simulated mean residence time, as discussed in 5.2.2. Among Pearson model and Han and Plummer model, the latter one showed considerable correlation (R2= 0.5) with the simulated median residence time. Figure 6.2-3 displays the correlation between model-simulated and hydrochemical model calculated transit time, excluding the negative values appeared in hydrochemical model calculations.

(a) (b)

(d) (c) Figure 6.2-3 Correlation between the hydrochemical models calculated and steady-state model-simulated residence times; (a) Pearson model vs simulated mean RT, (b) Pearson model vs simulated median RT, (c) Han and Plummer model vs simulated mean RT, (d) Han and Plummer model vs simulated median RT and the black dotted line in 1:1 line. Some of the apparent ages of the hydrochemical models showed a shift beyond the maximum range, Table 6.2-1. The flow path analysis (PMPATH) does not take into consideration the dispersion and diffusion of solute transportation. Whilst 14C can get affected by dissolution- fractionation processes from external sources. Hence 14C can be retarded, degraded or even some case have slightly different net transport direction than groundwater (Suckow, 2014). Thus, the complicated evolution of 14C in water could lead the deviation of tracer travel time beyond the modelled travel time boundary of idealised groundwater. Again, the area of per cell in the model was 0.25 square kilometres (500 m * 500 m). Therefore, the travel path distributions were very diverse (from local to regional flows) with high skewness. On the contrary, the samples were collected from a small portion of the well area in reality. Due to this difference in area, the comparison between two residence times will not be in perfect 1:1 ratio. Figure 6.2-4 and Table 6.2-1 show the comparison of residence times visually and numerically respectively.

dence time (Years) time dence Resi

Figure 6.2-4 Comparison of residence times after model simulation and hydrochemical model calculation

Table 6.2-1 Groundwater residence time determined by PMPATH and apparent age determined by 14C tracer; where “C” = contaminated or modern sample having negative values in the calculation

Residence Time (Years) Locations Median Mean value Acceptable Range Pearson Model Han and value from from simulated by Plummer model PMPATH PMPATH PMPATH PZ08 SC 701 952 340-2062 C C PZ14 SC 646 736 285-1709 1584 3459 PZ15 SC 1650 6334 619-47605 4418 6183 PZ16 F 684 860 19-9887 C 1148 Matola 1 1007 3144 32-28378 C C M1 232 319 155-704 1532 3385 M2 1092 1211 933-2153 1591 3370 M3 1336 1484 518-2952 5730 7448 M5 907 986 685-2143 462 2666

Though the absolute values of tracer modelled (both mean and median) and hydrochemical modelled residence times are different, they are in the same order of magnitude. Besides, the discrepancy between tracer residence time and modelled residence time are quite acceptable and explainable. Therefore, 14C tracer data affirms the uncertainties in the steady-state model parameters are within a considerable limit. And lastly, keeping the difficulties of 14C calculation in mind, it can be asserted that 14C the tracer age somewhat reflects the actual groundwater travel times of Matola wetlands.

6.3 Salinisation evolution

To test the hypotheses of this study, numerical groundwater modelling, SEAWAT, was initiated to evaluate the flushing of saltwater, from the aquitards on a geologic timescale of 5000 years. At the beginning, the sensitivity of effective porosity of aquitard clay layers were modelled to have the idea about the range of salinity variations. Over 5000 years, the variations in chloride concentration for two extreme cases of effective porosity was not more than 1000 mg/L in Paleo SEAWAT, Figure 5.2-6. For Modern SEAWAT, the range was even shorter, approximately 200 mg/L for 5000 years, Figure 5.2-9. The Paleo SEAWAT model had uncertainty about the initial source of salinity. Considering only paleo seawater as the primary source of chloride (19,000 mg/L) in some portions of upper aquitard, semi-confined aquifer and all over the lower aquitard, the model took approximately 3000-4000 years to achieve contemporary salinity distribution by flushing out the salt from both aquifers. Concurrently, when the initial source of chloride was alike present-day distribution, the model took 1500-2000 years to flush out the salt and produce freshwater condition (≤700 mg/L of chloride) in both aquifers. Despite having model parameter uncertainties, it can be claimed from the salinity evolution models that the significant portion of saltwater has already been flushed out from the aquifers, if the source is certainly connate marine water only. The models also correspond to the hydrochemical information, the ongoing freshening process, discussed in 0. Agreeing to the SEAWAT models, the freshening rate is constrained by the porosity or thickness of the aquitard layers. Nonetheless, Matola wetlands would turn into a freshwater environment in no sooner than next 1000-1500 years.

6.4 Salinisation Sources

The Paleo SEAWAT model and 14C calculations of this study further emphasised the results of Br/Cl ratio from (Nogueira, 2017). Since all the findings pointed towards the mid-Holocene apparent seawater intrusion and consequently entrapped seawater in the aquitards as the source of salinity in Matola wetlands; an attempt was taken to link the modelled and 14C calculated transit time with the geological evolution of the study area. Ramsay (1995) has studied the 9000 years of sea-level change along the coastline of South-Africa. He found that during mid- Holocene, several transgressions of sea-level occurred due to a combined effect of isostatic emergence and warmer ocean temperature, Figure 6.4-1. According to his research, sea-level reached its current level at 6500 BP. After that sea level ascended to +1.5 m above the current level at about 6000 BP and retained this level upto 5080 BP. At between 5080 - 4650 BP sea level rise kept on going again and reached to +2.75 m above present datum. Further escalation of sea-level in the mid-Holocene ended up being +3.5 m above the present level at around 4480 BP, and it was the highest still-stand of the South-African coast in the Holocene period. In post- Holocene period, a regression took place which lowered the sea-level to its present level at 3880 BP and further descend to -2 m at approximately 3000 BP. Afterwards, another sea-level upsurge to + 1.5 m at 1610 BP occured, and the coastline of Mozambique obtained its current stage at about 900 BP (Siesser, 1974).

Figure 6.4-1 Holocene sea-level fluctuations in southern Africa based on beach rock dating, Source: (Ramsay, 1995)

From the above discussion, it was inferred that the highest sea-level was +3.5 m at about 4000 BP. The maximum transit time of groundwater in the model was calculated to be in the range of 3000-4000 years mostly. Therefore, it was assumed that the origin of the aquitard salinity is the mid-Holocene marine transgression. Nevertheless, considering the coastal boundary remained constant from the Holocene period, a map was generated for +4 m sea-level rise with present DEM data showed inconsistency with the assumption, Figure 6.4-2 (a).

(a) (b) Figure 6.4-2 (a) Marine transgression of Holocene period (up to 4 m), (b) Relative elevation of sea-level needed to inundate the sampling points (60 m) The observation points seem too far away from the submerged area to be affected by the paleo- sea level rise. Besides, according to the hydraulic head gradient of this area, discussed in 3.5, groundwater flows from West to East and also rules out the possibility of saltwater transportation from coast to inland. Hence, another map was generated for investigating the relative elevation of sea-level needed to inundate area of the sampling points. The second map exhibited the sea-level increase as high as +60 m from the present condition, which is unattainable, Figure 6.4-2 (b). Accordingly, the repeated marine transgression-regression of Holocene period apparently cannot explain the geological origin of entrapped saltwater in the aquitards alone. At this stage, the assumption from Br/Cl ratio, studied by Nogueira (2017) was also reconsidered. Nogueira excluded the possibility of halite dissolution as one of the dominant salinisation mechanisms of the greater Maputo region. Generally, during the evaporation of water (irrespective of meteoric or seawater), the Br/Cl ratio remains consistent until halite gets fully saturated (Bottomley, et al 1994). Consequently, if evaporites are formed from complete evaporation of seawaters without being saturated and afterwards again completely get dissolved, the Br/Cl ratio will not be affected. Halite dissolution in some of the samples of this study lie marginally on the boundary of mixing range, hence halite dissolution as a source of salinity cannot be neglected totally, Figure 5.1-7. Therefore, along with the Holocene marine transgression theory, two additional possible origins of salinity were considered- (1) leaching of evaporitic strata; and (2) slow diffusion of formation water from the aquitard.

Subsurface salt dissolution is an important source of brines in the coastal sedimentary basin which increases with depth and temperature (Land, 1987). This salt might get dissolved with considerable recharge rates and subsequently get ejected into groundwater. Moreover, the regression of sea-level and subsequent evaporation might form a discontinuous layer of salt deposits within the sedimentary layers being formed. Thus leaching of evaporitic halite strata is also plausible owing to the fact that evaporation trend is quite high in the study area. In general, the connate salt water does not exist longer within the shallow subsurface due to the normal flushing of rainwater through time. Still the phreatic aquifer may have some pockets of residual saltwater on the places of low hydraulic gradients and hydraulic conductivities. Förster (1975) discussed the geological history of the sedimentary basin of southern Mozambique and showed several marine transgression-regression phases along with the changes of the paleo-coastlines. During these phases from the Cretaceous throughout the Quaternary period, some of paleo-coastlines reached more than 50 km inland from the present- day shoreline. Interestingly, the inferred coastlines pass very close to modern Matola River, Figure 6.4-3. Similar findings were conferred by Salman and Abdula (1995) with a detailed description of the evolution of the facies and the local lithology. According to Salman and Abdula (1995), Jurrasic deposits are absent in the coastal part of Mozambique Basin and occur only in the continental facies. The sedimentation of the coastal marine deposits start from the Early Cretaceous. This sequence of sedimentation is linked to the initial Gondwana break-up and a progressive invasion of the ocean along south-eastern border of Africa. Thus, there is a possibility of trapped connate water in the lower aquitard. This ancient saltwater might be coming up through the lower aquitard with a very slow diffusion rate and get mixed with comparatively modern water. This mixing might decrease the overall apparent travel time but increase the salinity of the groundwater. Findings from Förster (1975) along with Salman and Abdula (1995) thus fortifies the hypothesis of modern recharge water mixing with connate seawater.

Figure 6.4-3 Inferred positions of paleogeographic coastlines in southern Mozambique, red box indicates the present location of Matola River; Source: (Förster, 1975) Generally, 14C tracer is used for chronological events of past 45 ka (approximately), whereas 36Cl dating is useful to 100 ka–1 ma old water (IAEA, 2013). Besides, there are some limitations of 14C tracers which make the dating of ancient groundwater more complex. Therefore, 36Cl accompanying 14C tracer can be used to validate the travel time of groundwater of Matola wetlands. Besides, the reconstruction of a paleo-hydrogeological model with detailed marine transgression-regression phases will aid to minimise the uncertainties and reproduce the salinity evolution to the closest.

Conclusions and Recommendations

An integrated study of regional hydrogeochemistry, environmental isotopic analysis, and groundwater flow models have been undertaken to trace the saltwater residence time and origin together with the groundwater evolution of Matola wetlands to address the research objectives. The Piper plot diagram reveals that mixing between salinisation and freshwater followed by cation exchange is the prominent mechanism in the groundwater. As a result, Na-Cl type is the dominant hydrochemical facies along with limited Na-MIX / Mg-Cl facies in some portions of the wetlands. The Stuyfzand water classification shows that water samples vary from fresh to brackish-salt, though the majority of the samples have saltwater composition. The spatial distribution of salinities admits the chemical evolution of GW through the Stiff diagram plot. The salinities decrease from the Westside of the Matola River to the eastside which appears to be related to groundwater freshening potential. The latter is linked to land cover and different hydraulic conductivities which has also been confirmed by numerical modelling. The Stiff plot further shows that shallow water has comparatively more chloride concentration than the deep GW. Bivariate plots of major ions and ion ratios vs. chloride are conferred to observe the hydrochemical changes and to distinguish mixing processes of the water samples. Enrichment of Na/Cl in brackish groundwater are from salinisation and subsequent cation exchange, whereas the depletion of Na/Cl in fresher groundwater are due to flushing followed by cation exchange. The enrichment of Ca in saline groundwater and depletion in shallow groundwater suggests its ongoing reaction with calcite minerals and soil matrix. 18O vs 2H analysis provides information about climate conditions of the study area while recharging. Evaporation of shallow groundwater and surface water seems quite intense, as these waters are enriched in water stable isotopes. In the semi-confined aquifer the stable isotopes with more depleted values indicate a paleo water or a different climate condition during recharge. A conservative mixing model, 18O/Cl was used to interpret the salt origins. It implies possible mixing between fresh and saline GW, in some cases, halite dissolution might also take place. The transit time distributions (TTDs) were generated with the help of a steady-state particle- tracking MODFLOW-PMPATH model. The model was developed based on the existing calibrated groundwater flow model. The re-adjustment of the MODFLOW-PMPATH model was executed to understand the flow paths and travel time of Matola River, from upstream to downstream. Since insignificant variations were found after re-adjustment, this study stuck to the initial calibrated groundwater flow model. The manual sensitivity analysis of the steady- state model regarding the residence time of groundwater was also undertaken. Carrying this exercise reveals the most sensitivity towards the particle numbers, followed by effective porosity and hydraulic conductivities. The flow paths explain the mixing processes well, as the observation points in the model had both local and regional flows.

Considering the steady-state groundwater flow model as a basis, two saltwater transport models (using SEAWAT) were developed to demonstrate the evolution of the salinisation. Paleo SEAWAT model was used to test the hypothesis regarding the potential sources of saltwater of this study. On the other hand, Modern SEAWAT model was run to examine and predict the approximate freshening rate of Matola wetlands’ GW. The sensitivity analysis of SEAWAT models with respect to effective porosity of aquitard layers were considered prior to the result analysis. The result of Paleo SEAWAT model coincides with the hydrochemical analysis indicating the trapped seawater in the aquitard units during the-Holocene repeated marine transgressions as the probable source of the salinity. But according to present elevation and coastline of Maputo, the areal extent of the Holocene sea-level rise (+4 m) does not reach up to the newly sampling locations. Again, the result of Present SEAWAT model predicts the brackish-saline Matola wetlands to be a freshwater environment within next 1000-1500 years. However, if the slow diffusion of formation water is responsible for the salinity, it will take much more time than this expected time. 14C (DIC) tracer was used to date the groundwater travel time. 14C (DIC) is a sensitive tracer which gets affected by external sources easily. Therefore, six hydrochemical models namely- Tamers, Mook, Pearson, Fontes-Garnier, IAEA and Han-Plummer were used to consider all the possible hydrochemical reactions and to calculate the groundwater travel time as close as possible. Among six hydrochemical models, Pearson and Han-Plummer models had a considerable correlation with the numerical model simulated groundwater transit time. Despite several uncertainties involved in constraining the tracer age as well as the numerical model residence time calculations, the difference between these two residence times was marginal. Both of the residence times admits the presence of ancient seawater entrapped in the aquitard layers and the ongoing process of mixing with modern recharge water. The evaluation of the existing flow models was also accomplished with the aid of 14C results. Since two residence times were in the same order of magnitude, the numerical model seems to be a close approximation of the present-time Matola wetland condition. Notwithstanding the fact that the results of the hydrochemical analysis, the 14C results and the Paleo SEAWAT model agree about the ancient entrapped seawater, the geological history of the study area does not match appropriately with the findings. Thus the origin of the saltwater in the aquifer system is still difficult to trace, due to the ambiguity of the paleo-coastlines. So, in addition to the mid-Holocene marine transgressions, two other sources of saltwater are also considerable: leaching of evaporitic strata and slow diffusion of formation water from the aquitard. Due to the limitations of this research, several recommendations can be suggested for future studies to refine the understanding of the origin and evolution of saltwater in Matola aquifer system. If possible, boreholes needs to be purged approximately three times to the well volume prior to sample collection. Thus, the actual representativeness of the groundwater samples can be ensured. Samples for environmental isotope dating should be collected over long time. Although samples for 36Cl were collected with the intension of incorporating with 14C result, the lab analysis data are still expected and so could not be included with the present work. Therefore, multi-tracer dating approach is also recommended in future for validating the tracer age estimation. Field observation of halite occurrence is essential to support the findings of this study. The inversion of tracer concentration to derive a mean transit time with the help of lumped parameter models can be operated to calculate the travel time. Moreover, a paleo-

hydrogeological reconstruction together with variable-density groundwater flow and salt transport model is necessary to take into account the marine transgression and regression phases thoroughly over the past 60 ka. This model will help to solve the complexity of the origin and mechanisms of Matola wetlands salinisation and to take future management decision. Notwithstanding, the current results will provide a strong contribution to targeting correct adaptation and mitigation measures for groundwater salinity of the Matola wetlands.

References

Ameen, A. (2019) Sustainable development of groundwater resources under salinity and climate change impacts in Great Maputo , Mozambique. IHE-Delft.

Andreetta, R. (2018) Impact of climate change on groundwater recharge in the Great Maputo Aquifer, Mozambique. UNESCO- IHE.

Appelo, C.A.J. and Postma, D. (2005) Geochemistry, Groundwater and Pollution, 2nd Edition.

Atkinson, A. P. et al. (2014) ‘Using 14C and 3H to understand groundwater flow and recharge in an aquifer window’, Hydrol. Earth Syst. Sci. Copernicus Publications, 18(12), pp. 4951– 4964. doi: 10.5194/hess-18-4951-2014.

Ayadi, R. et al. (2018) ‘Hydrogeological and hydrochemical investigation of groundwater using environmental isotopes (18O, 2H, 3H, 14C) and chemical tracers: a case study of the intermediate aquifer, Sfax, southeastern Tunisia’, Hydrogeology Journal, 26(4), pp. 983–1007. doi: 10.1007/s10040-017-1702-1.

Bagheri, R., Bagheri, F. and Eggenkamp, H. G. M. (2017) ‘Origin of groundwater salinity in the Fasa Plain , Southern Iran , Hydrogeochemical and isotopic approaches’, Environmental Earth Sciences, 76(662), pp. 1–19. doi: https://doi.org/10.1007/s12665-017-6998-6.

Bahir, M., Ouhamdouch, S. and Carreira, P. M. (2018) ‘Geochemical and isotopic approach to decrypt the groundwater salinization origin of coastal aquifers from semi-arid areas (Essaouira basin, Western Morocco)’, Environmental Earth Sciences. Springer Berlin Heidelberg, 77(13). doi: 10.1007/s12665-018-7663-4.

Bottomley, D. J., Gregoire, D. C. and Raven, K. G. (1994) ‘Saline ground waters and brines in the Canadian Shield: Geochemical and isotopic evidence for a residual evaporite brine component’, Geochimica et Cosmochimica Acta, 58(5), pp. 1483–1498. doi: https://doi.org/10.1016/0016-7037(94)90551-7.

Boukhari, K. et al. (2015) ‘Origin of recharge and salinity and their role on management issues of a large alluvial aquifer system in the semi-arid Haouz plain, Morocco’, Environmental Earth Sciences, 73(10), pp. 6195–6212. doi: 10.1007/s12665-014-3844-y.

Carol, E., Kruse, E. and Mas-Pla, J. (2009) ‘Hydrochemical and isotopical evidence of ground water salinization processes on the coastal plain of Samborombón Bay, Argentina’, Journal of Hydrology, 365(3–4), pp. 335–345. doi: 10.1016/j.jhydrol.2008.11.041.

Cendón, D.I., Haldorsen, S., Chen, J., Hankin, S., Nogueira, G.E.H., Momade, F., Achimo, M., Muiuane, E., Mugabe J. and Stigter, T. Y. (2019) ‘Hydrogeochemical aquifer characterization and its implication for groundwater development in the Maputo district, Mozambique’, Quaternary International, pp. 1–14.

Cendón, D. I. et al. (2019) ‘Hydrogeochemical aquifer characterization and its implication for groundwater development in the Maputo district, Mozambique’, Quaternary International. Elsevier, pp. 1–14. doi: 10.1016/j.quaint.2019.06.024.

Chairuca, L. et al. (2018) Hydrogeology of Mozambique - Earthwise, Africa Groundwater Atlas: Hydrogeology of Mozambique. British Geological Survey.

Delsman, J. R. et al. (2014) ‘Paleo-modeling of coastal saltwater intrusion during the Holocene: An application to the Netherlands’, Hydrology and Earth System Sciences, 18(10), pp. 3891– 3905. doi: 10.5194/hess-18-3891-2014.

Eichinger, L. (1983) ‘A Contribution to the Interpretation of 14C Groundwater Ages Considering the Example of a Partially Confined Sandstone Aquifer’, Radiocarbon, 25(2), pp. 347–356.

FIPAG (2012) Estudos Ambientais e Sociais para o Sistema de Abastecimento de Água do Grande Maputo.

Fontes, J. ‐ C and Garnier, J. ‐ M (1979) ‘Determination of the initial 14C activity of the total dissolved carbon: A review of the existing models and a new approach’, Water Resources Research, 15(2), pp. 399–413. doi: 10.1029/WR015i002p00399.

Förster, R. (1975) ‘The geological history of the sedimentary basin of southern Mozambique, and some aspects of the origin of the Mozambique channel’, Palaeogeography, Palaeoclimatology, Palaeoecology, 17, pp. 267–287.

Freeze, R. Allan and Cherry, J. A. (1979) Groundwater. Available at: https://www.un- igrac.org/sites/default/files/resources/files/Groundwater book - English.pdf.

Le Gal La Salle, C. et al. (2013) ‘Origin of groundwater salinity (current seawater vs. saline deep water) in a coastal karst aquifer based on Sr and Cl isotopes. Case study of the La Clape massif (southern France)’, Applied Geochemistry. Elsevier Ltd, 37, pp. 212–227. doi: 10.1016/j.apgeochem.2013.07.006.

Gallagher, D. . et al. (2000) ‘Performance of models for radiocarbon dating of groundwater: An appraisal using selected Irish aquifers’, Radiocarbon, 42(2), pp. 235–248.

Geyh, M. A. (2000) ‘An overview of 14C analysis in the study of groundwater’, Radiocarbon, 42(1), pp. 99–114.

Godwin, H. (1962) ‘Half-life of radiocarbon’, Nature, 195, pp. 984–985.

Graas, S. and Savenije, H. H. G. (2010) ‘Salt intrusion in the Pungue estuary, Mozambique: effect of sand banks as a natural temporary salt intrusion barrier’, Hydrology and Earth System Sciences Discussions, 5(4), pp. 2523–2542. doi: 10.5194/hessd-5-2523-2008.

Han, D. et al. (2011) ‘Applied Geochemistry Geochemical and isotopic evidence for palaeo- seawater intrusion into the south coast aquifer of Laizhou Bay , China’, Applied Geochemistry. Elsevier Ltd, 26(5), pp. 863–883. doi: 10.1016/j.apgeochem.2011.02.007.

Han, D. M. et al. (2014) ‘Chemical and isotopic constraints on evolution of groundwater salinization in the coastal plain aquifer of Laizhou Bay , China’, JOURNAL OF HYDROLOGY. Elsevier B.V., 508, pp. 12–27. doi: 10.1016/j.jhydrol.2013.10.040.

Han, D., Post, V. E. A. and Song, X. (2015) ‘Groundwater salinization processes and reversibility of seawater intrusion in coastal carbonate aquifers’, Journal of Hydrology. Elsevier B.V., 531, pp. 1067–1080. doi: 10.1016/j.jhydrol.2015.11.013.

Han, L. F., Niel Plummer, L. and Aggarwal, P. (2014) ‘The curved 14C vs. δ13C relationship in dissolved inorganic carbon: A useful tool for groundwater age- and geochemical interpretations’, Chemical Geology. Elsevier B.V., 387, pp. 111–125. doi: 10.1016/j.chemgeo.2014.08.026.

Han, L. F. and Plummer, L. N. (2013) ‘Revision of Fontes & Garnier’s model for the initial 14C content of dissolved inorganic carbon used in groundwater dating’, Chemical Geology. The Authors, 351, pp. 105–114. doi: 10.1016/j.chemgeo.2013.05.011.

Han, L. F. and Plummer, L. N. (2016) ‘A review of single-sample-based models and other approaches for radiocarbon dating of dissolved inorganic carbon in groundwater’, Earth- Science Reviews. Elsevier B.V., 152, pp. 119–142. doi: 10.1016/j.earscirev.2015.11.004.

Han, L. F., Plummer, L. N. and Aggarwal, P. (2012) ‘A graphical method to evaluate predominant geochemical processes occurring in groundwater systems for radiocarbon dating’, Chemical Geology. Elsevier B.V., 318–319, pp. 88–112. doi: 10.1016/j.chemgeo.2012.05.004.

Hung Van Pham et al. (2019) ‘Paleo-hydrogeological reconstruction of the fresh-saline groundwater distribution in the Vietnamese Mekong Delta since the late Pleistocene’, Journal of Hydrology: Regional Studies. Elsevier, 23(July 2018), p. 100594. doi: 10.1016/j.ejrh.2019.100594.

IAEA (1993) ‘Mathematical models and their applications to isotope studies in groundwater, hydrology.’, Proceedings of a final research Co-ordination meeting held in Vienna, 1-4 June 1993, TECDOC-777(June), p. p 279.

IAEA (2013) Isotope Methods for Dating Old Groundwater. Vienna.

Kalin, R.M. and Long, A. (1993) ‘Application of hydrogeochemical modelling for validation of hydrologie flow modelling in the Tucson basin aquifer, Arizona, United States of America’, in Mathematical models and their applications to isotope studies in groundwater hydrology, pp. 147–178.

Karroum, M. et al. (2017) ‘Geochemical processes controlling groundwater quality under semi arid environment: A case study in central Morocco’, Science of the Total Environment, 609, pp.

1140–1151. doi: 10.1016/j.scitotenv.2017.07.199.

Kemper, K. E. (2004) ‘Groundwater-from development to management’, Hydrogeology Journal, 12(1), pp. 3–5. doi: 10.1007/s10040-003-0305-1.

Kim, Y. et al. (2003) ‘Hydrogeochemical and isotopic evidence of groundwater salinization in a coastal aquifer: a case study in Jeju volcanic island, Korea’, Journal of Hydrology. Elsevier, 270(3–4), pp. 282–294. doi: 10.1016/S0022-1694(02)00307-4.

Kinzelbach, W. and Simcore Software (2012) Processing Modflow: An Integrated Modeling Environment for the Simulation of Groundwater Flow, Transport and Reactive Processes, Simcore Software.

Land, L. S. (1987) ‘The major ion chemistry of saline brines in sedimentary basins’, in Banavar JR, Koplik J, W. K. (ed.) Physics and chemistry of porous media II; Ridgefield, Connecticut: AIP Conference Proceedings, 154, pp. 160–179. Available at: https://doi.org/10.1063/1.36392.

Larsen, F. et al. (2017) ‘Groundwater salinity influenced by Holocene seawater trapped in incised valleys in the Red River delta plain’, Nature Geoscience, 10(5), pp. 376–381. doi: 10.1038/ngeo2938.

Livshitz, Y. et al. (2005) ‘Sources of salinity and boron in the Gaza strip: Natural contaminant flow in the southern Mediterranean coastal aquifer’, Water Resources Research, 41(1), pp. 1– 20. doi: 10.1029/2004wr003344.

Maloszewski, P. and Zuber, A. (1996) ‘Manual on mathematical models in isotope hydrology’, International Atomic Energy Agency Technical Report …, p. 207. Available at: http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:Manual+on+mathematical+ models+in+isotope+hydrogeology#0.

McGuire, K. J. and McDonnell, J. J. (2006) ‘A review and evaluation of catchment transit time modeling’, Journal of Hydrology, 330(3–4), pp. 543–563. doi: 10.1016/j.jhydrol.2006.04.020.

Meredith, K. T. et al. (2018) ‘Evolution of dissolved inorganic carbon in groundwater recharged by cyclones and groundwater age estimations using the 14C statistical approach’, Geochimica et Cosmochimica Acta, 220, pp. 483–498. doi: 10.1016/j.gca.2017.09.011.

Modica, E.; Buxton, H. T. and Plummer, L. N. (1998) ‘Evaluating the source and residence times of groundwater seepage to streams, New Jersey Coastal Plain’, Water Resouces Research, 34(11), pp. 2797–2810.

Mongelli, G. et al. (2013) ‘Tracing groundwater salinization processes in coastal aquifers: a hydrogeochemical and isotopic approach in the Na-Cl brackish waters of northwestern Sardinia, Italy’, Hydrology and Earth System Sciences, 17(7), pp. 2917–2928. doi: 10.5194/hess-17-2917-2013.

Mook, W. G. and Plicht, J. Van Der (1999) ‘Reporting 14C activities and concentrations’, 74

Radiocarbon, 41(3), pp. 227–239.

Muiuane, E. (2007) The Quality of Groundwater in and around Maputo city, Mozambique. Maputo.

Müller, T. et al. (2016) ‘Use of multiple age tracers to estimate groundwater residence times and long-term recharge rates in arid southern Oman’, Applied Geochemistry. doi: 10.1016/j.apgeochem.2016.08.012.

N’da, B.; Bouchaou, L.; Reichert, B., Hanich, L.; Dann, S.; Ait Brahim, Y., Chehbouni, A.; Beraaouz, E. and Michelot, J. L. (2018) ‘Chemical Tracers and Stable Isotopes Mixing Models for Groundwater Quality and Recharge Study in the Moroccan High Atlas Mountains’, in Calvache M., Duque C., P.-V. D. (ed.) Groundwater and Global Change in the Western Mediterranean Area, pp. 235–244. doi: 10.1007/978-3-319-69356-9.

NCEA (2015) Climate Change Profile Mozambique.

Nogueira, G. et al. (2019) ‘Understanding groundwater salinization mechanisms to secure freshwater resources in the water-scarce city of Maputo, Mozambique’, Science of the Total Environment. Elsevier B.V., 661(December), pp. 723–736. doi: 10.1016/j.scitotenv.2018.12.343.

Nogueira, G. E. H. (2017) Tracing the Hydrochemical Water Types and Salinization Mechanisms in the Great Maputo Area as a Function of Groundwater Recharge, Hydrogeological Properties and Human Activities. UNESCO-IHE.

Oceans, U. A. of the (2019) Human Settlements on the Coast. Available at: http://www.oceansatlas.org/subtopic/en/c/114/ (Accessed: 18 March 2019).

Oszewski, Piotr M and Zuber, A. (1991) ‘Influence of matrix diffusion and exchange reactions on radiocarbon ages in fissured carbonate aquifers’, Water Resouces Research, 27(8), pp. 1937– 1945.

Ouhamdouch, S., Bahir, M. and Carreira, P. M. (2017) ‘Geochemical and Isotopic Tools to Deciphering the Origin of Mineralization of the Coastal Aquifer of Essaouira Basin, Morocco’, Procedia Earth and Planetary Science. The Author(s), 17(June 2018), pp. 73–76. doi: 10.1016/j.proeps.2016.12.038.

Parkhurst, D L and Appelo, C. A. J. (2013) ‘Description of Input and Examples for PHREEQC Version 3—A Computer Program for Speciation Batch-Reaction, One-Dimensional Transport, and Inverse Geochemical Calculations: U.S. Geological Survey Techniques and Methods’, in Groundwater Book 6, Modeling Techniques (Techniques and Methods. Amsterdam: U.S. Geological Survey. Available at: http://www.csoonline.com/article/220073/enterprises-and- employees-the-growth-of-distrust.

Payne, et al. (1983) Dating of Groundwater, Guidebook on Nuclear Techniques in Hydrology - Technical Report Series No. 91.

Pearson, F. J. JR and White, D. E. (1967) ‘Carbon 14 ages and flow rates of water in Carrizo Sand, Atascosa County, Texas’, Water Resources Research, 3(1), pp. 251–261. doi: 10.1029/WR003i001p00251.

Peché, M. (2012) Geology of the KaNyaka Barrier Island System, Maputo Bay, Mozambique. University of Johannesburg.

Petalas, C. P. and Diamantis, I. B. (1999) ‘Origin and distribution of saline groundwaters in the upper Miocene aquifer system, coastal Rhodope area, northeastern Greece’, Hydrogeology Journal, 7(3), pp. 305–316. doi: 10.1007/s100400050204.

Pollock, D. W. (1989) Documntation of Computer Programs to Compute and Display Pathlines Using Results from the U.S. Geological Survey Modular Three-Dimensional Finite-Difference Groundwater Flow Model: U.S. Geological Survey Open File Report 89-381.

Ramsay, P. J. (1995) ‘9000 Years of Sea-level Chnage Along The Southern African Coastline’, Quaternary International, 31(54), pp. 71–75. doi: 10.2307/25007923.

Rosário, D. L. (2016) Impacto da Proliferacáo de Furos Na Regiáo do Maputo. Maputo.

Salman, G. and Abdula, I. (1995) ‘Development of the Mozambique and Ruvuma sedimentary basins, offshore Mozambique’, Sedimentary Geology, 96(1–2), pp. 7–41. doi: 10.1016/0037- 0738(95)00125-R.

Santucci, L., Carol, E. and Kruse, E. (2016) ‘Identification of palaeo-seawater intrusion in groundwater using minor ions in a semi-confined aquifer of the Río de la Plata littoral (Argentina)’, Science of the Total Environment. Elsevier B.V., 566–567, pp. 1640–1648. doi: 10.1016/j.scitotenv.2016.06.066.

Shiklomanov, I.A. and Rodda, J. C. (2013) ‘World water resources at the beginning of the twenty-first century’, Choice Reviews Online, 41(07), pp. 41-4063-41–4063. doi: 10.5860/choice.41-4063.

Siesser, W. G. (1974) ‘Relict and Recent Beachrock from Southern Africa’, Geological Society of America Bulletin, 85(December), pp. 1849–1854.

Siesser, W. G. and Dingle, R. V (1981) ‘Tertiary Sea-Level Movements around Southern Africa’, The Journal of Geology. The University of Chicago Press, 89(4), pp. 523–536. Available at: http://www.jstor.org/stable/30085066.

Smidt, E.H., Chutumia, I. I., Haas, P. and Egidio, N. M. (1990) Groundwater Exploitation in the Area of Maputo, Mozambique.

Sola, F. et al. (2014) ‘Identification of a Holocene aquifer-lagoon system using hydrogeochemical data’, Quaternary Research (United States). University of Washington, 82(1), pp. 121–131. doi: 10.1016/j.yqres.2014.04.012.

Stigter, T.Y. et al. (1998) ‘A hydrogeological and hydrochemical explanation of the groundwater composition under irrigated land in a Mediterranean environment, Algarve, Portugal’, Journal of Hydrology. Elsevier, 208(3–4), pp. 262–279. doi: 10.1016/S0022- 1694(98)00168-1.

Stigter, T. Y. et al. (1998) ‘A hydrogeological and hydrochemical explanation of the groundwater composition under irrigated land in a Mediterranean environment, Algarve, Portugal’, Journal of Hydrology, 208(3–4), pp. 262–279. doi: 10.1016/S0022-1694(98)00168- 1.

Stigter, T. Y. et al. (2006) ‘Impact of the shift from groundwater to surface water irrigation on aquifer dynamics and hydrochemistry in a semi-arid region in the south of Portugal’, Agricultural Water Management, 85(1–2), pp. 121–132. doi: 10.1016/j.agwat.2006.04.004.

Stigter, T. Y. et al. (2014) ‘Comparative assessment of climate change and its impacts on three coastal aquifers in the Mediterranean’, Regional Environmental Change, 14, pp. 41–56. doi: 10.1007/s10113-012-0377-3.

Stigter, T. Y. et al. (2017) ‘Combined Assessment of Climate Change and Socio-Economic Development as Drivers of Freshwater Availability in the South of Portugal’, Water Resources Management, 31(2), pp. 609–628. doi: 10.1007/s11269-015-0994-y.

Stigter, T. Y., Ribeiro, L. and Carvalho Dill, A. M. . (2008) ‘Building factorial regression models to explain and predict nitrate concentrations in groundwater under agricultural land’, Journal of Hydrology, 357, pp. 42–56. doi: 10.1016/j.jhydrol.2008.05.009.

Stuyfzand, P. J. (1989) ‘A new hydrochemical classification of water types’, in Regional Characterization of Water Quality (Proceedings of the Baltimore Symposium, May 1989), pp. 88–98.

Suckow, A. (2014) ‘The age of groundwater - Definitions, models and why we do not need this term’, Applied Geochemistry. Elsevier Ltd, 50(May 2014), pp. 222–230. doi: 10.1016/j.apgeochem.2014.04.016.

Szabo, Z. et al. (1996) ‘Age dating of shallow groundwater with chlorofluorocarbons , tritium / helium 3 , and flow path analysis , southern New Jersey coastal plain’, Water Resouces Research, 32(4), pp. 1023–1038.

Tamers, M. A. (1975) ‘Validity of Radiocarbon Dates on Groundwater’, Geophysical Survey, 2, pp. 217–239.

Terway, J. L. (2012) ‘Isotopes in Groundwater Hydrology’, in Isotope Tracers in Catchment Hydrology. Harare: IAHS Publ. no. 144, pp. 203–246. doi: 10.1016/b978-0-444-81546- 0.50014-8.

Tóth, J. (1963) ‘A Theoretical Analysis of Groundwater Flow in Small Drainage Basins’, Journal of Geophysical Research, 68(16).

Trasviña, J. A. C. (2018) Assessing saltwater intrusion in the Great Maputo aquifer under natural conditions and human pressures with Numerical Models. UNESCO-IHE.

Travi, Y. et al. (2008) ‘Evolution of groundwater systems at the European coastline’, Geological Society, London, Special Publications, 189(1), pp. 289–311. doi: 10.1144/gsl.sp.2001.189.01.17.

Treidel, H. et al. (2011) ‘Beneath the surface of global change: Impacts of climate change on groundwater’, Journal of Hydrology. Elsevier B.V., 405(3–4), pp. 532–560. doi: 10.1016/j.jhydrol.2011.05.002.

WE Consult (2013) Development of a Groundwater Model for the Maputo Metropolitan Area Aquifer.

Werner, A. D. and Simmons, C. T. (2009) ‘Impact of sea-level rise on sea water intrusion in coastal aquifers’, Ground Water, 47(2), pp. 197–204. doi: 10.1111/j.1745-6584.2008.00535.x.

Wigley, T. M. L. (1975) ‘Carbon 14 Dating of Groundwater From Closed and Open Systems’, Water Resouces Research. Ontario, 11(2), pp. 324–328.

Xu, Y. S. et al. (2008) ‘The state of land subsidence and prediction approaches due to groundwater withdrawal in China’, Natural Hazards, 45(1), pp. 123–135. doi: 10.1007/s11069- 007-9168-4.

Yechieli, Y., Kafri, U. and Sivan, O. (2009) ‘The inter-relationship between coastal sub- aquifers and the Mediterranean Sea, deduced from radioactive isotopes analysis’, Hydrogeology Journal, 17(2), pp. 265–274. doi: 10.1007/s10040-008-0329-7.

Zhao, L. J. et al. (2018) ‘Origin and residence time of groundwater based on stable and radioactive isotopes in the Heihe River Basin, northwestern China’, Journal of Hydrology: Regional Studies. doi: 10.1016/j.ejrh.2018.05.002.

Zhou, Y. and Li, W. (2011) ‘A review of regional groundwater flow modeling’, Geoscience Frontiers. Elsevier B.V., 2(2), pp. 205–214. doi: 10.1016/j.gsf.2011.03.003.

Annexes

Annex I Steady-state Basic Model Configuration

. Basic package Groundwater flow model covers the greater Maputo region. The finite-difference grid represents the total area (13888 square kilometres) with each grid cell size of 500 m* 500 m, which results in 248 rows and 224 columns. Based on hydrogeology and lithological data, the total thickness of the aquifer system is considered to be 120 m. The aquifer system has four layers with variable layer thicknesses.

Table: Model layers with hydrogeological units

Hydrogeological Unit Model Layer

Unconfined/ Phreatic aquifer (F) 1 Aquitard (AQ) 2 Semi-confined aquifer (SC) 3 Semi-permeable unit (BT) 4

DEM (30 meters) data has been used to determine the elevations of each layer. To fill the no data sinks Kriging interpolation method was applied and later matched with the model grid resolution by up-scaling the elevations. The elevations are expressed in respect to surface elevation. Boundary packages Constant head (0 m) are given as cell status in layer 2 to 4 along the Indian Ocean boundary and the rest of the cells inside the boundary of the aquifer have active variable head. All the outer boundary cells have zero flux boundary (inactive) status. General head boundary (GHB) package has been utilised to replicate the Indian Ocean in the eastern and south-eastern boundary of the study area (Trasviña, 2018). . Layer property flow package (Trasviña, 2018) defined the horizontal hydraulic conductivity into 6 zones and Ameen (2019) recalibrated the steady-state model by improving the initial values of horizontal hydraulic conductivity of those zones. Vertical hydraulic conductivities were specified as anisotropy of 10 for all the model layers. Initial values and zoning were based on the hydrogeological map and previous studies conducted in the study area. . River package Incomati River is one of the main rivers in the study area which had river stage data from 11 hydrometric stations. Therefore, it is simulated in a river package in layer 1. A 30m DEM has been used for its digitisation. A value of 0.50m as the thickness of river bed sediments is considered in the model due to the unavailability of data. The hydraulic conductivity of the river bed is taken as 0.0075 m/d (Trasviña, 2018).

. Drain package No river stage data was available for the streams like Matola, Infulene, Chambadejovo and Cuenga Rivers. Hence, these streams / perennial rivers are added in the drain package in layer 1 to delineate the natural groundwater discharge zones (seepage and base flow). The bottom elevations of these streams are extracted from 30m DEM. The hydraulic conductivity of these channels ranges from 0.05 to 2 m/d (Trasviña, 2018). . Recharge package Trasviña (2018) specified 12 groundwater recharge zones and their respective recharge rates based on the earlier studies carried out by Mussa (2017) and (Nogueira, 2017). Later, the recharge rates were compared to the precipitation of the area, and the input recharge rate was determined. The recharge in the model is expressed in terms of the percentage rate as a result of running manual and PEST calibrations (Trasviña, 2018).

Annex II Description of parameters and notation used; formulas for isotopic fractionation

Source: (Han and Plummer, 2016)

Annex III Fig: Initial concentration of modern SEAWAT model; clockwise from upper left- Layer 1, Layer 2, Layer 3 and Layer 4

Fig: Initial concentration of paleo SEAWAT model; clockwise from upper left- Layer 1, Layer 2, Layer 3 and Layer 4

Annex IV Fig: Salinity distribution of Maputo aquifer after 5000 years (when simulated with present-day salinity distribution)

Layer 2 Layer 1

Layer 3 Layer 4

Annexe V

Table: Physiochemical parameter values, CO2 (g) as log CO2 pressure, ion concentrations (in mg/l), saturation indices (SI) for calcite, dolomite, gypsum and halite, electro neutrality balance (E.B.) and Stuyfzand water type of samples

pH Temp EC *DO CO2 *Cl- *HCO3- *SO4₂- *NO3- *Na+ *Mg₂+ *Ca₂+ *K+ *SiO2 SIcc SIdol SIgyp SIhal E.B. Stuyfzand (⁰C) ( µS/cm) (g) (%) Water Type PZ08 SC 6.5 27.6 1100 4.1 0.048 250 151.5 19 0.62 157.8 19.6 11 58.1 107 -1.73 -2.83 -3.13 -6 3.0 f2-NaCl+ PZ14 SC 6.9 26.2 10400 3.7 0.033 3000 370.5 329 45 1211 426 313 22.2 119 0.15 0.81 -1.17 -4.17 2.9 b3-NaCl- PZ15 SC 7.5 24.4 1515 2.1 0.010 232 325 66 0.15 204 21 39.1 7.5 35 0.05 0.17 -2.11 -5.92 -1.8 f3-NaMIX+ PZ16 F 7.2 26.6 29300 7.1 0.018 8200 443 1270 36 4280 730 420 155.9 47 0.46 1.56 -0.73 -3.25 1.2 b3-NaCl Matola 1 7.2 26 25200 2.5 0.011 8900 300 2050 34 2950 1020 1346 63.9 22 0.73 1.72 -0.12 -3.39 -3.1 b3-MgCl- Matola 2 7.8 26.4 20100 1.86 0.005 6600 443 540 28 2730 702 486 41.1 37 1.10 2.76 -0.97 -3.52 -0.8 b3-NaCl- M1 7.2 26.6 2800 3.7 0.021 550 396 87 6.4 386 68.4 100 10.1 69 0.22 0.64 -1.78 -5.31 7.5 B3-NaCl+ M2 7.1 25.9 5540 4.7 0.026 1320 367 161 30 740 121 153 18.0 89 0.12 0.50 -1.50 -4.69 3.4 b3-NaCl+ M3 7.1 24.6 6410 2.4 0.025 2180 420 338 36 936 181 117 32.6 39 0.04 0.61 -1.39 -4.42 -6.5 b3-NaCl- M4 6.8 26.4 1439 3.07 0.042 2020 307 104 37 612 311 451 32.3 76 0.11 0.43 -1.43 -4.62 7.9 b3-MgCl+ M5 7.3 25.4 7460 4.9 0.011 270 262 51 6 189 27.5 72.3 6.9 58 0.08 0.09 -1.99 -5.9 4.5 f3-NaCl+ Auger 4 7.3 28 24600 0.05 0.058 8400 1775 178 38 2810 990 848 63.5 38 1.46 3.40 -1.34 -3.43 -4.5 b5-MgCl-

Annexe VI Table: Lab analysis data of water stable isotopes

Lab ID Sample ID δ2H, in δ18O, in δ2H (GMWL) δ2H (LMWL) δ2H (Evaporation) ‰ ‰ W-4410 PZ08 SC -18 -3.87 -20.96 -18.169 -16.698 W-4411 PZ14 SC -20.5 -4.07 -22.56 -19.909 -17.778 W-4412 PZ15 SC -22.4 -4.51 -26.08 -23.737 -20.154 W-4413 PZ16 F -11.6 -2.71 -11.68 -8.077 -10.434 W-4414 M1 -21 -4.35 -24.8 -22.345 -19.29 W-4415 M2 -19 -4.09 -22.72 -20.083 -17.886 W-4416 M3 -18.4 -3.87 -20.96 -18.169 -16.698 W-4417 M4 -21 -4.27 -24.16 -21.649 -18.858 W-4418 M5 -21.8 -4.43 -25.44 -23.041 -19.722 W-4419 Matola 1 0.6 -1.19 0.48 5.147 -2.226 W-4420 Matola 2 8.8 0.93 17.44 23.591 9.222 W-4422 Auger 4 3.3 -0.15 8.8 14.195 3.39 W-4421 Auger 2 -16.4 -3.8 -20.4 -17.56 -16.32

Annexe VII Fig: Frequency distribution of transit time simulated by the steady-state model

Annexe VIII Fig: Cross-sections (column and row view) of the flow path simulated in PMWIN with exaggeration of 20. Blue colour represents phreatic aquifer, orange colour represents aquitard, cyan colour represents semi-confined aquifer, and yellow colour represents bottom aquitard

PZ08 SC

PZ14 SC

PZ16 F

PZ15 SC

M1 M2

M3 M5

Matola 1

All the locations